Sampling is an essential and extremely important sub-step in the whole measurement procedure, and can be the most relevant cause for measurement errors. Thus, sampling must strictly follow appropriate standards suited for the respective measurement task.
University of Kassel, Department for Sanitary and Environmental Engineering
(DESEE), Kurt-Wolters-Str. 3, 34125 Kassel/Germany, firstname.lastname@example.org
Competing interests: The author has declared that no competing interests exist.
Academic editor: Carlos N Díaz.
Content quality: This paper has been peer reviewed by at least two reviewers. See scientific committee here
Citation:F.-B. Frechen, 2014, Considerations on Sampling and measurement of odours (VDI 3880 & VDI 3885/1) and eNoses, Ist International Seminar of Odours in the Environment, Santiago, Chile, www.olores.org
Copyright: 2014 olores.org. Open Content Creative Commons license., It is allowed to download, reuse, reprint, modify, distribute, and/or copy articles in olores.org website, as long as the original authors and source are cited. No permission is required from the authors or the publishers.
Keyword: Odour sampling; static sampling; area sources; point sources; volume sources; odour emission capacity of liquids; electronic noses
Sampling is an essential and extremely important sub-step in the whole measurement procedure, and can be the most relevant cause for measurement errors. Thus, sampling must strictly follow appropriate standards suited for the respective measurement task. In the case of odour measurement, in Germany the VDI guideline 3880 “Olfactometry – Static sampling” (2011) and the VDI guideline 3885/1 “Olfactometry – Measurement of the odour emission capacity of liquids” (to be published 2014) are relevant. These guidelines describe the whole sampling process until the air samples are connected to the olfactometer. If, however, a measurement result is needed in real-time or near real-time, e.g. for process control purposes, it is not possible to use olfactometry. Instead, today arrays of multiple sensors, so called “electronic noses” (enoses) can be used after a mathematical evaluation is made in the respective case in order to establish a mathematical model that is not only able to explain the relationship but is also able to predict.
In the field of odour measurement, any measurement is preceded by the sampling procedure. Mistakes in sampling can massively attribute to measurement errors, thus, in order to assure quality of measurement, sampling has to be done according to sampling standards. In Germany, sampling standards can be found in many standards of DIN/VDI, and with respect to odour, besides other more general standards the most important ones are the VDI guideline 3880 (2011) and the VDI guideline3885/1 (to be published 2014).
VDI 3880 “Olfactometry – Static sampling” is the most important one, prescribing the sampling procedure for nearly every sampling situation one may encounter, like point, area and volume sources. This guideline deals with the interface between an odour emitting material and the ambient air or with odour already present in ambient air.
VDI 3885/1 “Olfactometry – Measurement of the odour emission capacity of liquids” is relevant in a situation where not the odour concentration per cubic meter of air, but instead the odour concentration per cubic meter of liquids is needed. This is of high relevance in all cases where liquids are the source of odour emissions, like domestic or industrial wastewater etc.
In cases where a measurement result is needed in real time for closed process control purposes, it is impossible to use olfactometry for measurement. In such cases one needs a real time (or “near real time”) measurement device, usually called an “electronic nose” or “enose”, rather than off line olfactometric measurements using test person panels etc.
This paper deals with the three aforementioned topics.
2. Sampling according to VDI 3880
The fundamental European standard on olfactometry, EN 13725:2003 (2003), does not give very precise instruction on sampling. Most of the – sparse – information is given in the informative annex J. However, it was felt by the European odour community that more stringent directions on sampling were utmost needed. Thus, the VDI guideline 3880 (2011) was elaborated and finally issued end 2011.
The guideline, besides the chapters on scope and terms & definitions, comprises chapters on
- “Planning of sampling and measurement”,
- “General requirements of sampling”,
- “Performance of sampling in relation of source type” and on
- “Quality assurance”.
The chapter on general requirements on sampling deals with several issues that partly were also dealt with in EN 13725 and further details some important issues that are marked in the following list:
- Working conditions
- Sampling equipment
- Sampling bag
- Performance of sampling for delayed olfactometry
- Sampling duration: 30 minutes
- Number of samples: min. 3 samples/source and operation condition
- Sample storage: < 6h, or stability has to be proven: 6 measurements (3 immediately after sampling, 3 after desired storage time, difference of more than a factor of 1.5 is not acceptable)
- Sample transport
2.2. Types of sources
Chapter “Performance of sampling in relation of source type” of the guideline covers all aspects of sampling at different types of sources. Figure 1 shows the different source types and compares with the EN 13725. As can be seen, the demarcation of active and passive sources is an important issue: » Odorants are emitted at the interface between the odour source and the free atmosphere. Depending on the nature of this interface, different sampling methods are necessary ".
Figure 1 : Different source types dealt with in VDI 3880, including numbering, compared with EN 13725:2003, (Frechen 2012a)
» If the interface is between a liquid phase and air (e.g. waste water, liquid manure) or between a solid phase and air (e.g. solid waste, compost), odorants can escape both by diffusion and as a result of aeration of the material (e.g. biofilter, pressurized activated sludge tank). If the waste gas’s emission velocity is significantly higher than the diffusion velocity caused by atmospheric diffusion, the source is termed an “active source”. If not, it is a “passive source”. "
» The dividing line between an active and a passive source is defined by convention as a flow velocity of 30 m/h as the arithmetic mean over the entire source interface. "
Concerning active point sources, only reference is made to DIN EN 15259 (2007).
Volume source are described as follows: » Volume sources have both a two-dimensional (horizontal) and a by no means negligible vertical extension. These are usually buildings or building complexes within which odorants are released. These odorants are subsequently discharged into the environment via gaps, windows, doors, roof openings, etc. at different heights and in different places. "
The basic approach to determine the odour flow rate emitted by volume sources is to representatively determine both the volumetric flow rate and the odour concentration. This can be done in many cases by adopting approaches familiar from the ventilation sector.
The most interesting and important sources are area sources. They can be discerned by being active or passive area sources according to the aforementioned criterion, which was set by convention. The decision on active or passive sources decides on the sampling equipment and sampling procedure to be used. Figure 2 (Frechen, 2012a) shows the criterion and typical values for aeration tanks in wastewater treatment plants.
Figure 2 : Demarcation of active and passive area sources according to VDI 3880, and examples of sources (Frechen 2012a)
2.3. Active area sources
The by far most common active area sources are biofilters. As can be seen from Figure 2, two sampling methods are basically possible: complete cover of the biofilter or point sampling using a sampling hood.
2.3.1. Complete cover
A complete cover is preferable to point sampling if the area is of a size that can be covered technically. Usually, days with very low wind are necessary.
The advantage of a complete cover is that the entire waste gas stream is emitted via one defined orifice and the determination of the total source strength is possible by measuring here, taking only three samples (see section 2.1). The total odour flow rate can thus be reliably determined.
Figure 3: complete cover at a small biofilter
2.3.2. Point sampling with a sampling hood
Point sampling of representative parts of large area sources is necessary if the area is too large to be totally covered. In this case, a sampling hood is used at a prescribed number of points on the area.
Figure 4 : left: principle of a sampling hood for active area sources (VDI 3880, 2011)
right: sampling hood in use. The automated sampler needed to achieve
a sampling time of 30 minutes is seen in the foreground (Frechen 2012a)
Genarally, this type of measurement is always beneficial if, in addition to the overall odour flow rate, a meaningful statement is required on the condition of the source (e.g. biofilter function) and its optimization. Only with point sampling inhomogeneous areas (flow velocity, odour concentration) can be detected, delivering a better picture of the state of individual part areas and regions of, e.g., a biofilter.
Thus, selection of the method depends upon the demands put on the respective measurement, and technical constraints.
Concerning the determination of sampling points VDI 3880 stipulates that » the area source is subdivided into part areas of the same size. The part areas are formed on the following principles:
- Rectangular sources are subdivided into a grid of part areas that are as square as possible.
- In the case of ground areas up to 100 m2, part areas are to be created of at least approximately 10 m2 each. At least four part areas are to be formed on ground areas < 40 m2.
- In the case of ground areas upwards of 100 m2 to 2 000 m2, at least one further part area is to be additionally created per roughly 100 m2.
For ground areas > 2 000 m2, 30 part areas are usually sufficient ."
For the sampling of part areas, a flow profile is determined before sampling in order to check the uniformity of flow and ascertain any irregularities. Odours are sampled with the sampling hood shown in Figure 4 in the part areas defined in accordance with then conditions cited above. Sampling in the sampling hood chimney may only take place after an at least five-fold change of air in the sampling hood.
Sampling time per point has to be constant in all part areas and amount to at least 3 min. For each single value (measurement result), the total sampling times in all part areas have to amount to at least 30 min.
For more details on calculation of the results on sources with homogeneous flow and with inhomogeneous flow, reference should be made to VDI 3880.
2.4. Passive area sources
Passive area sources are two-dimensional emission sources with a flow velocity of max. 30 m/h. Examples of passive sources are solid waste areas, sedimentation tanks of sewage treatment plants, slag beds, drainage channels, compost heaps (aerated and unaerated) and activated sludge tanks (aerated and unaerated areas, see Figure 2).
In Table 1 the requirements for the sampling hood are given. In Figure 5 an example of such a hood is drafted.
Table 1 : Requirements of the hood (VDI 3880, 2011)
Figure 5 : Example of a flow-through hood on a passive source (diagram of principle)
1 inlet fan; 2 activated carbon filter; 3 manometer; 4 outlet fan; 5 sampling port
6 diffuser plate (VDI 3880, 2011)
The number of measurement points is to be defined in accordance with the size and homogeneity of the passive area source. Homogeneous passive area sources (e.g. secondary clarifiers, water reservoirs etc.) with an area size of up to 100 m2 are to be sampled at at least three measurement points and those with an area size up to 1 000 m2 at at least five measurement points.
Sampling time is 30 min. For more details reference should be made to VDI 3880.
3. Sampling according to VDI 3885/1
Odour problems often arise from liquids that contain odorants. Depending upon circum stances like high turbulence etc., these odorants can be stripped off the liquid, generating nuisance in the environment. In order to assess a liquid concerning its ability to emit odorants, it is necessary to measure the total content of odorants in that liquid.
This is extremely important, as e.g. in a sewer with slow laminar flow, no big odorants mass transfer via the liquid-gaseous interface will happen even if the liquid itself already contains a lot of odorants. Assessing only the gaseous phase above this liquid would lead to a massive underestimation. If a point of high turbulence follows, odorants will easily be stripped off the liquid at that point, although the reason for the problem might be far away upstream.
In order to choose the right countermeasures at the right location e.g. in a sewer system, it is absolutely inevitable to know the quality of the liquid, i.e. measure the total amount of odorants in the liquid, because this determines the possible odorants release under e.g. conditions of high turbulence.
The procedure to attain this goal it the Odour Emission Capacity (OEC) measurement method. Frechen and Köster (1998) presented this method for the first time international in the year 1998, although the methodology was established by them since 1993. Frechen (2009) reported on the method and gave a summary of ten years of using this measurement method, doing more than 800 OEC measurements.
Also other researchers started to use the method, e.g. Zarra et al. (2011) who analysed 20 samples of 2 wwtps from March to December 2010.
Nevertheless, the method was not subject to any standardization so far. Thus, the German VDI initiated a committee to standardize this method as VDI guideline 3885 part 1 “Olfactometry – Measurement of the Odour Emission Potential of Liquids” (to be published 2014). The text is finished so far, and the issue of the draft for public comments is expected to happen very soon.
A 30 liter sample of the respective liquid is required. It is essential that sampling is done avoiding any turbulence as far as possible. Preferably the sampling conditions and the conditions at the sampling point and during sampling are documented by photos and a sampling protocol including facts like weather, air temperature, water temperature, etc.. Conductivity, pH-Value, oxygen concentration etc. should be scanned onsite. The test should be undertaken as soon as possible after sampling, as a change of the liquid, e.g. due to biological activity, must be avoided. Basically, the liquid sample is much less stable than the air samples generated during the OEC test.
3.2.2. OEC test reactor
A sketch of the OEC test reactor is shown in the left part of Figure 6.In the right part, three reactors, mounted in a trailer of DESEE, are shown.
Figure 6 : Sketch of the OEC test reactor (left). 3 reactors of DESEE fixed in trailer for on-site testing (right).
As most liquids are biologically active, it is essential to conduct the test immediately after sampling. This is why we at DESEE mounted all installations necessary for the test itself in a trailer to be able to conduct the OEC test onsite. Olfactometrc samples are much more stable than biologically active liquid samples.
The test reactor should be round to avoid unaerated corners. Usually, a standard fine bubble aerator well known from wastewater treatment plant aeration can be used to fill the total bottom area of the test reactor.
The test reactor has a closure head. Headspace volume should be less than 15 Liter. The closure has a connector so that sampling bags can be connected to withdraw samples of the off gas periodically.
Before conducting each single OEC test, the respective test reactor has to be filled with drinking water. Then, aeration has to be started with the same airflow that will be used in the subsequent test. After an aeration time of 5 minutes a sample has to be withdrawn and analyzed via olfactometry concerning the odour concentration cod.
The result should be below 100 ouE/m3. If not, the reason has to be identified and eliminated. A value of 100 ouE/m 3 or less is accepted as the inherent smell of the test setting, especially the fine bubble diffuser, which usually is made of EPDM or similar. Thus, it is accepted by convention that a signal (i.e. an odour concentration cod) is regarded as such if it is above 100 ouE/m3.
Immediately after sampling the sample should be filled into the test reactor avoiding any turbulence as far as possible. After filling, the test reactor is to be closed and aeration is switched on with the desired airflow.
From the measurements made by Frechen since 1994, the most common aeration flow rate is between 90 per hour and 100 per hour in relation to the sample volume of 30 Liter, resulting in an average airflow of 3,000 L/h. Although some tests were also undertaken with lower airflow down to below 10 1 /h, more than 93% were undertaken with an aeration flow rate between 90 1/h and 110 1/h.
After start of aeration, off gas samples are taken after fixed time intervals. In the tests undertaken by Frechen since 1994, the most common times were 1.5 min, 8 min, 16 min, 32 min and 60 min. By doing so, for each test 5 samples are generated, that have to be analyzed via olfactometry and can of course also be analyzed analytically. Usually, the sample volume is about 10 Liter and the sample bag filling time is below 20 s.
In the VDI 3885/1, test duration will be shortened. Total test time will be 32 minutes minimum, amount of aeration air will be 1,600 Liter minimum.
Also, the number of samples will be restricted to 4. This results in 5 samples per OEC test (1 with drinking water, see above) and allows for 3 OEC measurements per day and olfactometric team. Times for sampling are recommended to be 2, 6, 16 and 32 minutes, respectively.
Basis of the experimental data presented hereafter is the data pool measured by Frechen between April 1998 and June 2012, see Frechen (2012b). Older measurements undertaken in the years before 1998 were not taken into account. 835 OEC test were carried out in the period mentioned.
In general, the OEC is the area below the measurement curve and above a lower limit of integration which is accepted to be 100 ouE/m3 by convention, see section 3.2.3.
The basic formula of this integral is given in Figure 7. In fact the evaluation is done by simply calculating the area assuming a linear connection between the measured cod values.
The simplest case of possible results of an OEC test, hereafter called type a, is shown in Figure 7. In the course of the test, the measured values cod fall below 100 ouE/m3, which is accepted to be the lower limit of integration , see section 3.2.3.
Figure 7 : Evaluation of the test results type a
The OEC is the area between the measurement curve and the lower limit of integration, thus the sum of areas A1, A2 and A3. 10.2% of the total number of measurements mentioned were of type a.
Figure 8 shows the most likely type b of possible test results. 82.1% of the total number of measurements mentioned followed type b.
Figure 8 : Evaluation of the test results type b
In the course of the test, the measured values cod do not fall below 100 ouE/m3, (which becomes clear after olfactometric measurement, thus long after the test has ended). The last cod value measured, however, is lower than the value second to last. In this case, the OECmeasured is the area between the measurement curve and the lower limit of integration, thus the sum of areas A1 ... A5. The OECtotal is the OECmeasured plus the area A6 which is calculated simply by linear extrapolation in linear scale.
Accordingly, with type a, OECmeasured and OECtotal are identical.
An indicator of accuracy of the measurement might be the percentage of A6 related to the sum of areas A1...A6, thus to OECtotal, hereafter referred to as %OECcalc.
Besides these two types, there might be the result that OECtotal cannot be evaluated, see Figure 9. 7.7% of all tests were of type c.
In this case, no extrapolation is possible. Either the last two cod values are identical (1.5% of all tests), or the last value is even higher than the value second to last as shown in Figure 9 (6.2% of all tests).
In case of type c, no OECtotal can be calculated. However, OECmeasured can be calculated of course.
With test type c, it is not possible to give the OECtotal as the result of the test. In order to keep as much information as possible, today it is standard to give at least the OECmeasured and add the information whether the last two cod values were identical or the last cod value was even higher that the second to last value.
Table 2 gives an overview over types of test and evaluation possibilities.
Figure 9 : Evaluation of the test results type c
Table 2 : types of tests and possible results; number of cases per types
Type of test
% of all tests
>0% - <100%
Whenever OEC is mentioned without further specifications, OECtotal is meant. In general, it is recommended to report both the OEC total (if possible) and the OECmeasured. More details are given in Frechen (2014).
4. Use of enoses
Besides cost, one of the most important disadvantes of olfactometric measurement is that it is a typical off line measurement. If a real time or near-real time measurement is needed – and this usually is the case with closed loop process control activities – the sensoric method cannot be used.
A very important application of the OEC measurement is a closed loop process control of dosing chemicals into a sewer system in order to prevent odour nuisance, but also with the aim of minimization of chemical cost. In Paris, e.g., between 3 and 7 million Euro are spent each year only for chemicals that suppress unpleasant odours from the sewer system. Introduction of a closed loop dosing process control system with the aim of always dosing enough but never overdose chemicals could save huge sums.
This requires two things: an automated OEC test machine and an odour measurement using any available real time device. This device providing a signal indicating “odour” is what is called a multisensor array or “electronic nose” (enose).
Automation of the OEC test is more or less an electrical/mechanical task and was resolved at DESEE as early as 2005. So far we built automated OEC devices including enoses as sensor for odour in the cases of the new emscher sewer, a test unit for Berlin water services and recently for the city of Hannover. All devices use two reactors in parallel, one for sulphide measurement and the other for OEC measurement. In Figure 10, the right photo shows the two test reactors.
Figure 10 : automated OEC device for the emscher sewer, Frechen (2012a)
In the case of Berlin water services, the reactors, shown in Figure 11, had to be larger, as here 4 enoses in parallel were tested, see Rouault et al. (2013).
Figure 11 : automated OEC device for research at the Berlin water services. Note the much larger reactors that are connected to 4 enoses in parallel. See Frechen (2012a) and Roualt et al. (2013)
It is aimed in all these devices to produce two signals: sulphide concentration and OEC of the wastewater. Frechen and Giebel (2011, 2014) described in detail the need for enoses and the basic mathematical procedures, because mathematical evaluation is one crucial step.
4.2. Mathematical evaluation
In order to obtain information concerning odour from the raw data of a multisensor array with 4 to 10 different sensors, it is necessary first to get the enose known in order to work with it properly. Then, as much parallel measurement between olfactometry with test persons (as the anchor value) and enoses are necessary before it makes sense to work on the data under a mathematical point of view.
Intense research activities resulted in the fact that today DESEE’s datapool holding parallel measurement results of enoses and olfactometric measurement from the same sample is as big as can be seen in Table 3, presumably representing the world’s largest data pool of parallel measurement between enoses and olfactometry. This is essential when dealing with mathematical models to represent “odour” by the signals of a multisensory array.
Table 3 : DESEE’s data pool as of end 2012, Frechen (2012a)
Once dealing with mathematical evaluation of enose vs. olfactometric data, it is important to carefully observe the following possible sources of errors:
- type of eNose (batch-like measurement or continuous reading)
- sampling location – different constituents of odours
- Unsynchronized clocks of sampling devices involved (this might be overcome by doing the eNose measurement from the same sample bag as the olfactometric measurement)
- Sample bags storage time
- Uncertainty of olfactometric measurement
- It has to be differentiated between THE REAL ODOUR and THE ODOUR MEASURED BY OLFACTOMETRY
- Olfactometry measurement has an error itself. According to Boeker (2005) this can be estimated to be 50%.
- This means: if an eNose perfectly explains the odour as measured by olfactometry, (claimed by some vendors of eNoses: R2 = 1 !!), the model has an error of 50%!
- The eNose itself
- no information provided by the manufacturers
- t90-time differs from sensor to sensor
- t90-time differs from odour to odour
- Some sensor signals do not converge to a final value at all – then there is no
t90-time at all…
- oscillating readings (but mean can be used)
- uniformity of devices (do 2 eNoses of the same type and age deliver the same readings?)
- aging of sensors/enoses
- The odour’s nature
- Although already well established in several (specific) industrial processes, eNoses still have problems with the very varying composition of odours – e.g. wastewater odours, but also others. This is why every application should be preceded by a measurement program for model calibration
- The mathematical evaluation itself
- Selection of data to be included
- Type of model
- Model preconditions
Concerning the mathematical evaluation, two approaches are commonly used today:
- Qualification (“which kind of smell is that”)
- Pattern recognition with
- Configuration frequency analysis
- Shape analysis
- Neural networks
- Pattern recognition with
- Quantification (“ouE/m3”)
- Partial Least Square PLS
- Linear Regression
- Nonlinear Regression
- Logistic Regression (below or above a given limit value)
- Neuronal networks (able to achieve nearly perfect explanation, but ability for prediction is crucial, see below!)
4.3. Explanation or prediction?
Most important with mathematical models is to distinct between
- explanation (done in nearly 99% of cases) and
- prediction capabilities determination by data splitting!
If vendors claim that their eNose is capable of measuring odour with a correlation coefficient of near 1, they usually talk about explanation of historical data (ouE/m3 measured using olfactometry vs. ouE/m3 calculated from the eNose raw data). There are no vendors known that really check for prediction abilities. If so, they should testify that they
- used the bootstrap-procedure: derive a model for explanation basing on part (e.g. 50%) of the data (training data) (this is the explanation part, possibly with high correlation coefficients), and
- applied this model to the other part of the data for prediction.
Correlation coefficients near 1 for prediction are not achievable as set out before (REAL odour vs OLFACTOMETRIC odour).
In practice, usually the data set for bootstrap purposes is divided into two data sub-sets of the same size, one (training data set) to derive a mathematical model, the other to test the model derived. A sample result is given in Figure 12.
Figure 12 : example of the bootstrap method for prediction ability determination, Frechen and Giebel (2011)
Dataset 1 was the training data set, the outcome of the model is demonstrated with linear regression (green line, R2=0,9707). Applying the model derived to the test data set (Dataset 2) resulted in the red dots and the red line – it is obvious that the model derived was good for evaluation (and thus also good for advertising), but prediction ability (what will the next measurement bring?) was poor.
The VDI 3880 on sampling will further reduce variability of measurement results due to a more strict and uniform boundary conditions prescription.
The VDI 3885/1 on OEC from liquids will make this method available, offering solutions to a broad range of so far unresolved problems with odours at all plants where liquids are relevant.
With enoses, a crucial issue is whether only explanation is done (which is nice but insufficient) or whetherprediction ability was tested. Determining prediction ability is the only method that makes sense when judging on the usability of a specific enose brand.
Boeker, P. (2005). Methodology of Odour Measurements. Unpublished report
EN 13725:2003-04 (2003) “Air quality – Determination of odour concentration by dynamic olfactometry”
EN 15259:2007 (2007) “Air quality – Measurement of stationary source emissions; Requirements of measurement sections and sites and of the measurement objective, plan and report”.
Frechen, F.-B. (2009) 10 Years of OEC measurement: Methodology, application, results and future development. in: Wasser - Abwasser - Umwelt. Schriftenreihe des Fachgebiets Siedlungswasserwirtschaft, Volume 31. Kassel University Press. 2009. ISBN 978-3-89958-608-4
Frechen, F.-B. (2012a) Curso “Control de Olores en Plantas de Tratamiento de Aguas servidas y Residuos industriales liquidos” – Seminar on Odours, Santiago, Chile, 28.08.2012.
Organized by ecotec
Frechen, F.-B. (2012b) Evaluation of OEC data for preparation of the new VDI Guideline 3885/1 CEt Chemical Engineering transactions , Vol. 30, pp. 19-24, 2012. ISBN 978-88-95608-21-1; ISSN 1974-9791; DOI:10.3303/CET1230004
Frechen, F.-B. (in press, 2014) Odour measurement in liquids – the odour emission capacity OEC. in: Handbook of Environmental Odour Management, edited by Frechen, F.-B.; Stuetz, R.; van Harreveld, T.; Guillot, J.M. (in press, to be published 2014)
Frechen, F.-B., Giebel, S.M. (2011) The new VDI guideline 3885/1 for OEC measurement and why we might need electronic noses. in: Proceedings of PETrA 2011 - Pollution and Environment - Treatment of Air, 17. - 19.05.2011, Prag. ISBN 978-80-02-02293-0
Frechen, F.-B., Giebel, S.M. (in press, 2014) Mathematical evaluation of raw data from electronic noses.
in: Handbook of Environmental Odour Management, edited by Frechen, F.-B.; Stuetz, R.; van Harreveld, T.; Guillot, J.M. (in press, to be published 2014)
Frechen, F.-B.; Köster, W. (1998) Odour Emission Capacity of Wastewaters – Standardization of Measurement Method and Application. Water Science and Technology, Vol. 38, No. 3, pp. 61-69, 1998, IAWQ, Elsevier Science Ltd., Oxford, Great Britain, ISBN 0 08 043391 X
Rouault, P.; Schwarzböck, T.; Frey, M.; Giebel, S.; Frechen, F.-B. (2013) Multigas-sensor systems for sewer odour measurement - Evaluation of four different E-noses based on tests under realistic conditions. Paper for presentation at the 7th International Conference on Sewer Processes & Networks, Sheffield, United Kingdom, 28th to 30th August 2013
VDI guideline 3880 (2011) “Olfactometry – Static sampling”. Beuth Verlag GmbH, 10772 Berlin. © Verein Deutscher Ingenieure e.V., Düsseldorf 2011
VDI guideline 3885 part 1 (to be published 2014) “Olfactometry – Measurement of the odour emission capacity of liquids”. © Verein Deutscher Ingenieure e.V., Düsseldorf
Zarra T., Giuliani S., Naddeo V., Belgiorno V. (2010) Control Of Odour Emission In Wastewater Treatment Plants By Direct And Undirected Measurement Of Odour Emission Capacity. Proceedings of the 4th IWA Conference on Odours and VOCs, 17. - 21.10.2011, Vitoria, Brazil