You are on page 1of 10

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/233408941

An easy uncertainty evaluation of the COD titrimetric analysis in correlation


with quality control and validation data. Method applicability region

Article  in  Analytical Methods · October 2012


DOI: 10.1039/C2AY25687H

CITATIONS READS

14 3,504

5 authors, including:

Elisavet Amanatidou Eleni Trikoilidou


University of Western Macedonia Technological Educational Institute of Western Macedonia
84 PUBLICATIONS   276 CITATIONS    24 PUBLICATIONS   309 CITATIONS   

SEE PROFILE SEE PROFILE

Georgios Samiotis Nikolas-Plutarch Benetis


University of Western Macedonia National Hellenic Research Foundation
30 PUBLICATIONS   202 CITATIONS    86 PUBLICATIONS   1,476 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

A special issue of Resources (ISSN 2079-9276), entitled:"Advanced and Modern Processes of Wastewater Treatment". View project

Improving water management and supply infrastructure via smart technologies, policies and tools, SAVE - WATER View project

All content following this page was uploaded by Georgios Samiotis on 27 March 2014.

The user has requested enhancement of the downloaded file.


Analytical
Methods
PAPER

An easy uncertainty evaluation of the COD titrimetric


analysis in correlation with quality control and
Cite this: Anal. Methods, 2012, 4,
4204 validation data. Method applicability region
Elisavet Amanatidou,* Eleni Trikoilidou, Georgios Samiotis, Nikolas- P. Benetis
and Nikolaos Taousanidis

Chemical Oxygen Demand (COD) estimates the amount of organic oxidizable matter in wastewaters. It is a
critical parameter for assessing the quality of effluents prior to discharge, in the design and operation of
wastewater treatment plants, and in environmental modelling and monitoring. In this study COD was
determined by a closed reflux titrimetric method according to 5220C, APHA, 21st edn, 2005. This
method is applicable to COD values between 40 and 400 mg L1 and chloride concentrations up to
2000 mg L1. Chloride interference is a serious problem in COD measurement for wastewaters
containing low organic matter concentrations. The reliability of the results depends on a proper
validation of the measurement procedure that includes the calculation and evaluation of the
measurement uncertainty. This study develops a practical, understandable and simple way for
calculating and assessing the Combined Standard Uncertainty (CSU) of COD measurement, mainly based
on already existing quality control and validation data. The CSU assessment is based on the existing
literature trueness and precision target limits, without any additional assumptions. Diagrams correlating
the trueness of the COD measurement mean values with different chloride concentrations and the
Received 12th April 2012
Accepted 25th October 2012
laboratory control charts’ limits offer a tool for improving the performance of the determination and
define the method application area. The validation data that were obtained in our laboratory
DOI: 10.1039/c2ay25687h
(accredited according to ISO 17025) and the proposed methodology for CSU calculation were verified in
www.rsc.org/methods the frame of intra and inter-laboratory measurements.

1 Introduction containing low organic matter and high chloride concentra-


tions. There are a number of analytical methods for the deter-
Chemical Oxygen Demand (COD) is dened as the amount of a mination of COD. In this study the COD closed reux titrimetric
specied oxidant that reacts with the sample under controlled method according to 5220C, Standard Methods for the Exami-
conditions. Both the organic and inorganic components of a nation of Water and Wastewater, APHA, 21st edn, 20052 was
sample are subjected to oxidation, but in most cases the organic used by the Laboratory of Environmental Chemistry and
components predominate and are of greater interest. COD is Wastewater Treatment (accredited according to ISO 17025).
one of the most widely used parameters in the eld of waste- This method is applicable to COD values between 40 and
water analysis and is frequently used as an indicator of waste- 400 mg L1 and chloride concentrations up to 2000 mg L1.
water pollution with organic compounds. The determination of The principle of the closed reux titrimetric COD measure-
COD in wastewaters is of highest importance for operation ment method is the oxidation of the organic matter by a boiling
control of industrial and municipal wastewater treatment mixture of a known excess of 0.01667 M potassium dichromate
plants, wastewater effluent monitoring, etc. It requires, (K2Cr2O7) and concentrated sulfuric acid in the presence of a
however, a detailed study of the presence of interfering species catalyst (Ag+ salt). Aer digestion, the remaining unreduced
that can signicantly affect the estimation of the target oxid- K2Cr2O7 is titrated with ferrous ammonium sulfate (FAS),
ability of the sample.1 Fe(NH4)2(SO4)2$6H2O, to determine the amount of K2Cr2O7
The most common interference in COD measurement is due consumed and the oxidizable matter is calculated in terms of
to coexisting chloride ions, especially for wastewaters oxygen equivalents.
Oxidation of most organic compounds occurs to 95 to 100%
of the theoretical value. The most common interference is
Technological Educational Institution of West Macedonia, Department of Pollution
Control Technologies, Environmental Chemistry and Wastewater Treatment
coming from chloride ions. Chloride reacts with silver ions to
Laboratory, Koila, Kozani 50100, Greece. E-mail: eamanatidou@kozani.teikoz.gr; precipitate silver chloride and thus inhibits the catalytic activity
Fax: +30 24610 39682; Tel: +30 24610 68015 of silver. Such interferences are negative tending to restrict the

4204 | Anal. Methods, 2012, 4, 4204–4212 This journal is ª The Royal Society of Chemistry 2012
Paper Analytical Methods

oxidizing action of the dichromate ion. Under the digestion Viana da Silva et al. (2011) evaluated the magnitude of the
procedure applied in COD analysis, chlorides react with estimated measurement uncertainty by comparing the relative
dichromate ions to produce chlorine and chromic ions, causing or absolute expanded measurement uncertainty with dened
a positive interference, eqn (1). target values.5
The main objective of the present study is to develop a
6Cl + Cr2O72 + 14H+ / 2Cr3+ + 3Cl2 + 7H2O (1) practical, understandable and simple way for calculating and
assessing the measurement uncertainty of COD, mainly based
The most oen used standard methods for COD determina- on the already existing quality control and validation data. The
tion are suitable for samples containing less than 2000 mg L1 absence of established measurement uncertainty target limits
Cl ions. This limitation renders the COD determination inap- led this laboratory to develop a methodology for the Combined
plicable for brackish and sea water which have a concentration of Standard Uncertainty assessment based on the existing true-
chloride ions around 0.55–1.37 g kg1 and 19.2 g kg1 respec- ness and precision target limits, without any additional
tively. More complicated, expensive and time consuming tech- assumptions. Moreover, the laboratory identied the method
niques designed to measure COD in saline waters are available. application area of COD measurement, by correlating uncer-
Difficulties that are caused by the presence of chloride ions can tainty, trueness, precision and laboratory applied control limits,
be overcome largely by complexing with mercuric sulfate (1 g with and without chloride interference. The proposed meth-
HgSO4 per 50 mL sample), in the Hg : Cl ratio of 10 : 1, up to odology is an alternative way for ensuring the application of the
2000 mg L1 Cl,2 before the reuxing procedure (closed reux method under controlled conditions.
titrimetric method) that leads to formation of a soluble salt or a
soluble unreactive complex (chloride masking).2,3
HgSO4 reacts with Cl and forms the easily soluble HgCl2 or
complexes of type HgCl42 are formed (eqn (2) and (3)) in the 2 Material and methods
sample. The extent to which HgCl2 in aqueous solutions is ionized The aim of this study is to (a) develop an easily applied method
is small. The Hg2+ concentration in saturated HgCl2 solution is for the uncertainty evaluation and assessment of COD titri-
108 g ion per L, which is less than the H+ concentration in water metric measurement useful for both the academic and the
(covalent bond). Thus the presence of sulfuric acid in the solution industrial sector and (b) develop a way for determining safe
does not give replacement reaction of chloride. control chart limits taking into account the interferences under
The reactions of chloride masking are shown in eqn (2) and (3). the condition of acceptable accuracy and precision. This
approach was examined by method validation and verication
HgSO4 + 2Cl / HgCl2 + SO42 (2)
and both trueness and precision as well as repeatability and
reproducibility (from intra and inter-laboratory comparisons)
HgSO4 + 4Cl / HgCl42 + SO42 (3) estimation.6
Trueness is the closeness of agreement between the average
Many other approaches that attempt to manage the problem of an innite number of replicate measured quantity values and
of chloride ion interference including addition of silver salts to a reference quantity value.7 The terminology is very similar to
mask chloride ions are utilized. For example, addition of that used in accuracy, but trueness applies to the average value
chromium(III) to lower the oxidation potential of the reaction in of a large number of measurements.8
eqn (1), determination of the amount of chloride ions oxidized Precision is the closeness of agreement between indications
by iodometric titration followed by a subsequent correction for or measured quantity values obtained by replicate measure-
chloride ions, oxidation with permanganate, etc.3,4 In the ments on the same or similar objects under specied condi-
oxidation of the organic load by potassium permanganate tions. The ‘specied conditions’ can be, for example,
KMnO4 for the COD measurement (Japanese Mn-COD method), repeatability conditions of measurement, intermediate preci-
the COD is known as the permanganate value. Compared to the sion conditions of measurement, or reproducibility conditions
Mn-COD, the Cr-COD method that is used as the standard in of measurement.7 Precision increases as random variations
USA and Europe has a higher detection limit and severe chlo- decrease. It is possible to have results, which are precise but not
ride ion interference because of the stronger oxidant and the accurate. They may be all close to a mean or average value but
longer digestion time.4 because of some systematic effect there is a bias rendering them
The reliability of the results depends on a proper validation all much higher (or lower) than the true value.8
of the measurement procedure that includes the calculation Trueness and precision alone may omit some uncertainty
and evaluation of the measurement uncertainty. sources. For this reason it is necessary to compare trueness and
There have been reported many different approaches of how precision computation (results) with target limits.
to calculate and assess the measurement uncertainty. Drolc For minimization of the error propagation, a controlled
et al. (2003) identied the major uncertainty sources and experimental procedure was applied, proposed recently by this
compared the combined uncertainty with the experimentally laboratory as “Errors Propagation Break-up (ERB)”.9 The
determined variation. The results showed that the major sour- Combined Standard Uncertainty (CSU) was calculated accord-
ces of uncertainty arose from repeatability at a high concen- ing to the error propagation law.10 The controlled experimental
tration level and volumetric steps at a low concentration level.1 procedure can be applied by considering a measurement

This journal is ª The Royal Society of Chemistry 2012 Anal. Methods, 2012, 4, 4204–4212 | 4205
Analytical Methods Paper

procedure as a series of discrete operations each of which may 2.1.1 IDENTIFICATION OF UNCERTAINTY COMPONENTS IN COD
be assessed separately. TITRIMETRIC DETERMINATION. In the closed reux titrimetric
The method validation and verication was achieved COD measurement method, the possible sources of error that
considering the repeated analysis data of in-house standards, as constitute the Combined Standard Uncertainty (CSU) compo-
well as inter-laboratory comparison programmes, both being nents are presented in Table 1. Some of them were considered
useful in establishing method precision. signicant and were taken into account in the CSU
The results of the inter-laboratory measurements conrmed calculation.
the assumptions and veried the calculations of the Combined Because the in-house or eld sampling stages do not form
Standard Uncertainty at the 95% condence level. part of the specied procedure, the components of uncertainty
The procedure steps involved during the present investiga- related to the sampling procedure were not included in this
tion were: calculation procedure.
(i) Evaluation of Combined Standard Uncertainty (CSU) 2.1.2 CALCULATION OF UNCERTAINTY COMPONENTS. The
mathematical expression. measurement procedure is considered as a series of discrete
(ii) Calculation of trueness and precision. operations. Each of them may be assessed separately to obtain
(iii) Comparison of calculated trueness and precision with estimates of the uncertainty associated with it.
standard target limits. Accordingly, as is described below, the separate uncer-
(iv) Calculation of type A uncertainty if trueness and preci- tainties for individual or groups of components were estimated
sion are acceptable. and expressed as standard uncertainty that contributed to the
(v) CSU calculation. CSU calculation.
(vi) Method application area determination, with the The digestion and the standard solution preparation
contribution of trueness, precision and control limits in one uncertainty were included in the repeatability (r)/intermediate
composite control chart and in correlation with the inuence of precision7 (within laboratory reproducibility, Rw)11 and it
the interfering substance. was not taken into account in calculating the Combined
The key point in this procedure is to bring trueness and Standard Uncertainty (CSU). Under controlled conditions of
precision within the target limits. Only under this condition the the method application, the total bias is insignicant
other steps can be applied. In this way the laboratory uses compared to the combined uncertainty and it can be
international literature accepted target limits to assess CSU and neglected in the total uncertainty calculation. In the
at the same time to ensure the measurement accuracy. following is presented the uncertainty calculation of the main
components as well as the CSU of the closed reux titri-
metric COD determination.9
2.1 Uncertainty calculation of COD titrimetric
All the assumptions were conrmed by repeated intra- and
determination
inter-laboratory measurements.9–12
The evaluation of the Combined Standard Uncertainty (CSU) is Sample volume uncertainty. The volume of the solution con-
a statistical approach of measurement, where the different tained in the volumetric ask is subjected to three major
uncertainty sources are estimated and combined into a single sources of uncertainty9,10,13–15
value. The basis for the estimation of measurement uncertainty (a) the volumetric glassware uncertainty (tolerance);
to be proposed is derived from the existing data; no special (b) the repeatability (glassware lling) uncertainty, u(R); and,
investigation is required from the laboratories. (c) the uncertainty caused by the volume difference between
For the uncertainty estimation the steps involved are: (a) the glassware calibration temperature (20  C) and the solution
measurand specication, (b) uncertainty source identication, temperature,
(c) uncertainty component quantication, and (d) CSU
calculation. u(t) ¼ DV/O3 and DV ¼ V20  aw  DT

Table 1 Uncertainty components in titrimetric COD measurements

Uncertainty components Uncertainty error sources Data sources

1. Sample volume - Filling - Glassware calibration certicate


- Temperature - Measurements
- Tolerance - Literature data
- Reading
2. Random phenomena (type A uncertainty) - Repeatability - Measurements
- Intermediate precision
3. Titration - Burette technical data - Burette calibration certicate
4. Weighing - Balance technical data (included in standard solution preparation uncertainty) - Balance calibration certicate
5. Standard solutions preparation - Weight, dilutions (included in type A uncertainty and in bias) - Measurements
6. Reagents purity - Included in standard solution preparation uncertainty - Reagent certicates of analysis
7. Method and laboratory total bias - Standard solutions - Certicates of analysis
- Measurements

4206 | Anal. Methods, 2012, 4, 4204–4212 This journal is ª The Royal Society of Chemistry 2012
Paper Analytical Methods

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where V20: glassware volume at 20  C, aw: volume expansion
Sr 2
coefficient, DT: temperature variation. uðbiasÞ ¼ uCRM 2 þ (7)
n
By a, b, and c, the combined sample volume uncertainty,
u(V), is calculated in eqn (4): where uCRM ¼ a/2 (a ¼ certicate value), Sr ¼ standard
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  2 deviation.
tol DV
uðV Þ ¼ pffiffiffi þ½uðRÞ þ pffiffiffi 2
(4) If the contribution of the method and laboratory total bias
3 3
uncertainty is insignicant, it is not taken into account in the
However, the glassware lling uncertainty13 is included in Combined Standard Uncertainty calculation and eqn (6) is
the measurement of the repeatability or/and the intermediate reformed to eqn (8):
precision (type A uncertainty), as well as in the standard solu- sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
    2ffi
u Vsample 2 uðV titration Þ
tion preparation uncertainty and it cannot be taken into uðcÞ ¼ c  þðRSDrÞ2 þ (8)
Vsample Vtitration
account in the volume uncertainty estimation. So the nal
relation is expressed in eqn (5): This equation can be applied safely in all titrimetric
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  2 determinations.
tol DV
uðV Þ ¼ pffiffiffi þ pffiffiffi (5) The contribution of the sensitivity coefficient has been
3 3
included in the estimation of component uncertainties and was
This way of calculation can be used also for automatic considered equal to unity.6
pipettes and dispensers. For instruments of variable volume, The expanded uncertainty, for a coverage factor k, is evalu-
the determination of the tolerance and the lling DV in the ated from the relationship: U ¼ ku(c). For an approximate level
closer volume of the measurement are necessary. of condence of 95%, k ¼ 2 and U ¼ 2u(c).
Titration uncertainty. Similarly, in the titration uncertainty are
involved the above three steps. 3 Results and discussion
Type A uncertainty. The estimation of type A uncertainty,10 was
For the purpose of this study, solution standards of 20, 40, 200
performed under repeatability and intermediate precision
and 400 mg L1 COD with and without the presence of chloride
conditions and expressed as RSDr or RSDRw. Because of the
were measured by the closed reux titrimetric method, as
Errors Propagation Break-up (ERB) procedure applied by this
control samples. The COD standard solutions were prepared by
laboratory, the use of one of the above type A uncertainty
suitable dilutions of potassium hydrogen phthalate stock
expressions did not cause any important differentiation from
solution (KHP). The presence of chloride was ensured by addi-
the total uncertainty calculation, as was conrmed by the rele-
vant intra- and inter-laboratory measurements. The ERB tion of NaCl in the solution stock. The inuence of the dilution
procedure considers that the measurement is a series of from the solution stock was controlled by repeated COD
measurements of diluted standards and comparison with the
discrete operations whose uncertainties can be assessed sepa-
laboratory control charts limits. The concentrations of the COD
rately. Furthermore the ERB imposes specic limitations to
standard solutions used for this study are given in Table 2.
some of these operations depending on their contribution in
For the measurements, a high purity reagents calibrated
the total uncertainty. This is a controlled experimental proce-
Brand automatic burette, a calibrated Hamilton soop
dure in the frame of the bias source elimination and was
dispenser, a calibrated Brand Transferpette S pipette, a cali-
recently presented by this laboratory.9
Combined Standard Uncertainty calculation. Taking into brated WTW CR2200 digester as well as a calibrated Mettler
account the uncertainty components mentioned, the Combined Toledo XS205 balance and glassware were used. The prepara-
tion of the standard solutions had signicant contribution in
Standard Uncertainty (CSU) of closed reux titrimetric COD
the estimation of the control sample concentration uncertainty.
determination is expressed by eqn (6).1,8,9
According to the ERB procedure, this laboratory calculated
uðcÞ ¼ c
separately the concentration uncertainties of the standard
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
    2  2ffi solutions and only standard solutions with uncertainty #1–2%
u Vsample 2 2 uðVtitration Þ uðbiasÞ were used for achieving the CSU management in the framework
 þðRSDrÞ þ þ (6)
Vsample Vtitration c0 of the internal quality control applied.8 The use of such
where u(Vsample), sample volume uncertainty; RSDr, type A
uncertainty expressed by the repeatability or the intermediate
Table 2 Concentrations of standard COD solutions
precision; u(Vtitration), burette volume uncertainty and u(bias),
method and laboratory total bias.9 The method and laboratory Concentrations of
total bias9 is estimated by the analysis of the Certicated standard solutions
Reference Material (CRM) under repeatability conditions. It is mg L1 COD Chloride concentrations mg L1 Cl
the combination of the standard deviation of the mean for n
20 1000 2000 3000 — — —
determinations and of the standard uncertainty of the certied 40 1000 2000 2500 3000 — —
value of the CRM, (at a level of condence of 95%). The total 200 1000 2500 3000 4000 5000 6000
bias7 is expressed in eqn (7): 400 2000 2500 3000 4000 5000 6000 7000 8000

This journal is ª The Royal Society of Chemistry 2012 Anal. Methods, 2012, 4, 4204–4212 | 4207
Analytical Methods Paper

standard solutions contributed in reducing the type A uncer- that all the experimental data, more than 2000 COD measure-
tainty of this study in which standard solutions were used as ments were made over a period longer than four years.
control samples. Control charts for the internal quality control Calculating trueness, the agreement among the average
were used.16 values that are obtained by a number of measurements is esti-
The titration was performed with 0.025 M ferrous ammo- mated, as a parameter of the method systematic errors. True-
nium sulfate (FAS) solution, suitable for low COD samples, ness is expressed by eqn (9):
instead of 0.1 M FAS proposed in Standard Methods for the
Examination of Water and Wastewater for most samples. Trueness ¼ 100  (Xmean  m)/m (9)
The FAS solution was standardized before each set of
measurements. If the FAS concentration was #0.024 M, a new where m is the true value.
reagent was prepared and standardized, in the frame of the ERB Depending on the measurand concentration, the trueness is
process applied. acceptable in comparison with the limits from Table 3.17
The trueness, the precision and the uncertainty of the COD If trueness is acceptable, precision, which is a parameter of
measurements in the above four concentration levels, with and the method random errors, is calculated as RSDr, and
without the presence of chloride, were calculated and compared compared with RSDrmax as given in Table 4 (ref. 18), for the level
with the relative target limits (Tables 3 and 4). of concentration m.
To obtain means and standard deviations, 16 samples were Precision is expressed by eqn (10).18
processed for each COD concentration level. It should be noted
RSDr ¼ Sr/Xmean (10)

where Sr is the standard deviation under repeatability


Table 3 Trueness target limits 17 conditions.
If RSDr # RSDrmax, the method precision is acceptable in the
Concentration of analyte 100  (Xmean  m)/m concentration level examined with reference materials (in-
house standards).
m < 1 mg kg1 50 to +20%
1 mg kg1 < m < 10 mg kg1 30 to +10%
In Tables 5 and 6 are presented the calculation of trueness,
m > 10 mg kg1 20 to +10% precision and standard deviation results of COD under repeat-
ability conditions for a large number of measurements in four
concentration levels, with and without the presence of chloride.
Table 4 Relationship of concentration and relative standard deviation (RSDr) Trueness and precision of COD control sample measurements
under repeatability conditions18 without the presence of chloride are acceptable for all the
concentration levels (Table 5).
Concentration of analyte
The variation of the trueness relative to chloride concentra-
Average fraction of Corresponding tion, for all the COD concentration levels, is shown in Fig. 1a–d.
analyte in the sample concentration of the analyte RSDr maximuma In this gure are clearly shown the experimentally determined
limits of chloride concentrations in which the trueness is
109 to 108 1–10 mg kg1 30%
acceptable (trueness limits: 20 up to +10%).
108 to 107 10–100 mg kg1 20%
107 to 106 100 mg kg1–1 mg kg1 15%
Under the presence of chloride ions, trueness and precision
106 to 105 kg1–10 mg kg1 10% of COD low concentration control samples, measurements are
105 to 104 10 mg kg1–100 mg kg1 7.5% acceptable. For high COD concentrations, trueness and preci-
104 to 103 100 mg kg1–1 g kg1 5.0% sion are acceptable for higher chloride ion concentrations than
103 to 102 0.1–1% 3.5%
the original method prescription implies (Table 6).
102 to 101 1–10% 2.5%
101 to 10 10–100% 2.0%
More specically, for 200 mg L1 COD and chloride concen-
trations greater than 3300 mg L1 Cl, the trueness is not
a
These values of relative standard deviation RSDr were obtained from acceptable because of the increasing systematic errors, while the
large experimental calculations of HORWITZ and based on these
results and the related graph. (Horwitz trumpet–funnel of Horwitz). precision is acceptable because the random phenomena are under
They are independent of the analyte’s nature, the substrate (matrix) of control. For 400 mg L1 COD, the trueness and the precision are
the analyzed material and the method of analysis. acceptable for chloride concentrations up to <8000 mg L1 Cl.

Table 5 Trueness, precision and standard deviation of COD measurements in four concentration levels, without the presence of chloride

Trueness
COD mg L1 Measurement average Standard deviation (SD) (20 up to +10%) Precision Observations

20 21.70 2.7 8.5% 3.1% (up to 7.5%) Trueness and precision: A


40 39.3 4 1.75% 2.65% (up to 7.5%) Trueness and precision: A
200 202.25 4.2 1.12% 2% (up to 5%) Trueness and precision: A
400 393 7.5 1.85% 1.92% (up to 5%) Trueness and precision: A

4208 | Anal. Methods, 2012, 4, 4204–4212 This journal is ª The Royal Society of Chemistry 2012
Paper Analytical Methods

Table 6 Trueness, precision and standard deviation of COD measurements in four concentration levels, with the presence of chloridea

Measurement Trueness
COD mg L1 Cl mg L1 average (20 up to +10%) Precision

20 1000 21.9 9.50 A 7.53 A (up to 7.5%)


2000 33.33 66.63 NA 5.39 A (up to 7.5%)
3000 56.39 181.94 NA 18.32 NA (up to 7.5%)
40 1000 40.26 0.65 A 3.1 A (up to 7.5%)
2000 43.41 8.52 A 2.44 A (up to 7.5%)
2500 51.8 29.5 NA 13.83 NA (up to 7.5%)
3000 61.07 52.7 NA 16.39 NA (up to 7.5%)
200 1000 202.02 1.01 A 1.97 A (up to 5%)
2500 209.025 4.98 A 4.97 A (up to 5%)
3000 214.64 7.32 A 4.57 A (up to 5%)
4000 230.44 15.22 NA 2.30 A (up to 5%)
5000 271.99 35.99 NA 1.35 A (up to 5%)
6000 288.72 44.36 NA 1.35 A (up to 5%)
400 2000 395.50 1.125 A 0.93 A (up to 5%)
2500 399.31 0.17 A 1.3 A (up to 5%)
3000 399.39 0.15 A 0.9 A (up to 5%)
4000 382.30 4.42 A 2.77 A (up to 5%)
5000 391.89 2.03 A 0.68 A (up to 5%)
6000 399.35 0.16 A 1.84 A (up to 5%)
7000 426.16 6.54 A 0.53 A (up to 5%)
8000 446.13 11.53 NA 0.22 A (up to 5%)
a
A: acceptable, NA: not acceptable.

Fig. 1 Trueness variation relative to chloride concentration for four different COD concentrations (a) 20 mg L1 COD, (b) 40 mg L1 COD, (c) 200 mg L1 COD and (d)
400 mg L1 COD.

This journal is ª The Royal Society of Chemistry 2012 Anal. Methods, 2012, 4, 4204–4212 | 4209
Analytical Methods Paper

Table 7 Uncertainty of COD measurements in four concentration levels without the presence of chloridea

COD mg L1  U
1 1
COD mg L  U COD mg L  U (calculated by using repeatability as
(calculated by using repeatability as (calculated by using intermediate type A uncertainty, including total
COD mg L1 type A uncertainty) precision as type A uncertainty) laboratory bias)

20 21.70  5.4 21.5  6.4 21.7  5.6


40 39.30  8 — 39.3  8.3
200 202.25  8.5 199.3  8.6 202.25  8.7
400 393  15.2 394.1  18 393  15.3
a
U ¼ ku(c). For condence level 95%, k ¼ 2 and U ¼ 2u(c).

Table 8 Uncertainty of COD measurements in four concentration levels with the presence of chloride

COD concentrations (mg L1) 20 mg L1 Cl 1000 2000 3000


COD mean value mg L1 20.75 21.9 56.39
U 6.63 2.4 20.66
40 mg L1 Cl 1000 2000 2500 3000
COD mean value mg L1 40.26 43.41 51.8 61.07
U 7 2.2 27.95 20
200 mg L1 Cl 1000 2500 3000 4000 5000 6000
COD mean value mg L1 201.85 209.025 214.64 230.44 271.99 288.72
U 7.52 16.77 20 5.33 8.90 8.00
400 mg L1 Cl 2000 2500 3000 4000 5000 6000 7000 8000
COD mean value mg L1 395.50 399.31 399.39 382.30 391.89 399.35 426.16 446.13
U 7.49 10.93 7.64 21.30 5.60 14.80 4.85 2.6

In Tables 7 and 8 are presented the Combined Standard Breakup (ERB) method,9 this laboratory contributed to reducing
Uncertainty (CSU) results for a large number of COD measure- the type A uncertainty even under repeatability conditions. In
ments at four concentration levels, with and without the pres- applying the ERB methodology the consideration of either
ence of chloride. This laboratory examined the contribution of repeatability or intermediate precision in the total uncertainty
the repeatability and intermediate precision in the CSU, by calculation did not alter signicantly the results, as presented in
calculating the type A uncertainty with repeatability and intra- Table 7.
laboratory intermediate precision data, at the selected concen- For example, in 200 mg L1 COD the uncertainty calculated
trations. Normally, the repeatability standard deviation, Sr, by using repeatability as type A uncertainty is 8.5 mg L1 and by
excludes major uncertainty contributions and is not suitable for using intermediate precision as type A uncertainty is 8.6 mg
the uncertainty estimation. Therefore, for the estimation of the L1. Because of the analytic accuracy improvement of the COD
measurement uncertainty, the intermediate precision standard testing by this laboratory through the use of rigorous quality
deviation SRw is used most oen. However, by applying a control and CSU management programs, the method and
controlled experimental procedure and a strict internal quality laboratory total bias is not signicant and is not specied as a
control for CSU management, i.e. the Error Propagation separate performance characteristic in CSU calculation. Its

Fig. 2 Laboratory control charts (mean value  2S) at 200 mg L1 COD (a) without the presence of chloride and (b) with 2500 mg L1 Cl.

4210 | Anal. Methods, 2012, 4, 4204–4212 This journal is ª The Royal Society of Chemistry 2012
Paper Analytical Methods

contribution in CSU is within the range 0.02% to 0.45% and it is


negligible as shown in Table 7;19–21 (0.045%, 0.33%, 0.045%,
0.022% for COD concentrations of 20, 40, 200 and 400 mg L1
respectively). For 200 mg L1 COD the uncertainty calculated by
using repeatability as type A uncertainty, including the total
laboratory bias, is 8.7 mg L1.
From Tables 7 and 8 is also observed that by increasing the
chloride ion concentration the COD measurement mean
values also increase, especially for the COD concentrations
that can tolerate lower chloride concentrations. At the same
time, the trueness and precision is also increasing (Table 6).
When trueness and precision (or one of them) becomes
unacceptable, the measurement is imprecise, although the
uncertainty may not change signicantly. The uncertainty in
most cases, specied solely as an isolated value can be
considered acceptable, but arriving at a decision only from
this assessment is wrong.22 For this reason, it is necessary to
correlate the trueness with the laboratory control charts’ limits
under the condition of acceptable precision. The laboratory
control charts, which demonstrate that the measurement Fig. 4 Method application area (gray shadow) for 400 mg L1 COD relative to
chloride concentration, trueness and control limits from the control chart con-
procedure performs within given limits, can be constructed
structed with chloride-free solution standard. The % values are the trueness
with solution standard in the presence of interfering corresponding to different COD mean values.
substances. In Fig. 2a are presented the laboratory control
charts at 200 mg L1 COD without chlorides and with control
For control samples of 200 mg L1 COD, chloride-free, the
limits 190–208 mg L1 COD. In Fig. 2b are presented the
method is applicable for samples with chloride concentrations
laboratory control charts at 200 mg L1 COD, with 2500 mg
<2500 mg L1 Cl (Fig. 3a) instead of 2000 mg L1 Cl proposed
L1 Cl and control limits 192.4–225.65 mg L1 COD; in this
from the standard method. Also, for control samples of 200 mg
case the laboratory control limits are expanded. This control
L1 COD with 2500 mg L1 Cl, the method is applicable for
charts can be applicable under the constraints of an accept-
able trueness and precision.
The correlations of the trueness of the COD measurement
Table 9 Results of COD inter-laboratory measurements
mean values with different chloride concentrations and the
laboratory control charts’ limits are presented graphically in Reference Laboratory
Fig. 3 and 4. Such diagrams, which can become useful tools for Date value Measurement z-score code
any laboratory, dene the method application area (gray
June–July 2008 96.2 97.2 0.1 026
region). The application area is limited from the lower and
April–May 2009 135 133 0.4 035
upper laboratory control limits and the acceptable trueness of October–November 2009 147 141 0.1 063
the COD measurement mean values in different chloride Dec–January 2010 144 151 0.5 046
concentrations. It can be implemented by laboratories in the Feb–March 2010 295 280 0.5 021
frame of analytical methods validation and internal quality June–July 2010 114 110 0.4 014
June–July 2010 34.9 35.57 0.8 061
control programs.

Fig. 3 Method application region (gray shadow) for control samples of 200 mg L1 COD relative to chloride concentration, trueness and control limits; (a) control
limits from the control chart constructed with chloride-free solution standard; (b) control limits from the control chart constructed with 2500 mg L1 Cl solution
standard. The % values are the trueness corresponding to different COD mean values.

This journal is ª The Royal Society of Chemistry 2012 Anal. Methods, 2012, 4, 4204–4212 | 4211
Analytical Methods Paper

samples with chloride concentrations <3300 mg L1 Cl References


(Fig. 3b). For control samples of 400 mg L1 COD, chloride-free
and control limits 376–411 mg L1 COD, the method is appli- 1 A. Drolc, M. Cotman and M. Roš, Accredit. Qual. Assur., 2003,
cable for samples with chloride concentrations <6400 mg L1 8, 138–145.
(Fig. 4). In the case of using control samples with chlorides the 2 Standard Methods for the Examination of Water and
control limits will be expanded and the method can be appli- Wastewater, APHA, Washington, 21st edn, 2005.
cable for samples with chloride concentrations <8000 mg L1 3 B. Vaidya, S. W. Watson, S. J. Coldiron and M. D. Porter, Anal.
because trueness and precision are acceptable up to this Chim. Acta, 1997, 357(1–2), 167–175.
concentration (Table 6). 4 S.-C. Chen, J.-H. Tzeng, Y. Tien and L.-F. Wu, Anal. Sci., 2001,
Therefore, COD measurements of urban and most industrial 17, 551.
wastewater, with such chloride concentration levels,23 can be 5 M. E. Viana da Silva, et al., Anal. Chim. Acta, 2011, 699, 161–169.
measured with satisfactory accuracy. Samples of higher 6 General Requirements for the Competence of Testing and
chloride concentrations can be measured by making suitable Calibration Laboratories, ISO-IEC 17025:2005, ISO, Geneva, 2005.
dilutions which minimize the interference. 7 JCGM, International Vocabulary of Metrology-Basic and
The lower COD quantitation limit, that can be determined General Concepts and Associated Terms, 3rd edn, 2012.
with acceptable trueness and precision as well as acceptable 8 Accuracy (Trueness and Precision) of Measurement Methods and
level of uncertainty, is 20 mg L1 COD for relative clean waters Results - Part 1: General Principles and Denitions, ISO 5725-1,
with chloride ion concentration up to 1000 mg L1 Cl. This 1st edn, 1994.
COD concentration is lower than the method relative limit 9 E. Amanatidou, E. Trikoilidou, L. Tsikritzis and F. Katsiouli,
(Tables 5–8). Talanta, 2011, 85, 2385–2390.
The results of the inter-laboratory measurements conrmed 10 Quantifying Uncertainty in Analytical Measurement,
the assumptions and veried the calculations of the Combined EURACHEM/CITAC Guide CG 4, ed. S. L. R. Ellison, M.
Standard Uncertainty.24 Rosslein and A. Williams, 2nd edn, 2000.
The inter-laboratory measurements were implemented 11 B. Magnuson, H. Hovit, M. Kryssel and T. Naykii., Nordest
according to LEAP Scheme Chemistry for wastewater. The Handbook for Calculation of Measurement Uncertainty Based
results, as z-score, for the examined parameter are presented in on Quality Control and Method Validation. Proceedings of 1st
Table 9. The z-score is considered satisfactory for values International Prociency Testing Conference, Romania, 2007,
|z| # 2. pp. 26–30.
12 M. Thompson, S. Ellison and R. Wood, Pure Appl. Chem.,
2002, 74, 835–855.
4 Conclusions 13 LAB 15, UKAS, London, 1st edn, 2002.
14 R. Lawn and E. Prichard, LGC/VAM/2000/108, 2000.
An easily applied method for the uncertainty evaluation of 15 B. Wamper and M. Rössleina, Talanta, 2009, 78(1), 113–119.
closed reux titrimetric COD measurement, applicable in all 16 M. Thompson and R. Wood, Pure Appl. Chem., 1995, 67, 649–666.
titrimetric determinations, useful for both the academic and 17 United Kingdom Accreditation Service, UKAS, London, 1997.
the industrial sector, was developed in this study. 18 Food Law Enforcement Practitioners, FLEP, 1995.
Laboratory control charts, demonstrating the measurement 19 The Fitness for Purpose of Analytical Methods. A Laboratory
procedure within specic limits can be constructed with solu- Guide to Method Validation and Related Topics, EURACHEM
tion standards in the presence of interfering substances, having Guide, EURACHEM, UK, 1st edn, 1998.
a matrix similar to the samples. In this case the laboratory 20 K. F. Yee, J. Autom. Chem., 1998, 10(3), 144.
control limits are expanded. Diagrams correlating the trueness 21 D. B. Hibbert, J. Chromatogr., A, 2007, 1158, 25–32.
of the COD measurement mean values with different chloride 22 P. Armishaw, B. King and R. G. Millar, Achieving traceability
concentrations and the laboratory control charts’ limits offer a in chemical measurement – a metrological approach to
tool for improving the performance of the determination and prociency testing, Accredit. Qual. Assur., 2003, 8(184), 190.
denes the method application area. It can be applied by 23 A. Riga, K. Soutsas, K. Ntampegliotis, V. Karayannis and
laboratories as a way to achieve a good internal quality control G. Papapolymerou, Desalination, 2007, 211, 72–86.
(QC) involved in this type of chemical determinations. 24 F. Raposo, et al., TrAC, Trends Anal. Chem., 2010, 29(9).

4212 | Anal. Methods, 2012, 4, 4204–4212 This journal is ª The Royal Society of Chemistry 2012

View publication stats

You might also like