Abstract. 1 Introduction

Transcription

1 Transactions on Ecology and the Environment vol 8, 996 WIT Press, ISSN 7- Combining reversed modelling and odour detection campaigns to determine the odour emission from a landfill for domestic waste C. Mensink, G. Cosemans, M. Wevers, R. De Fre, P. Geuzens Vlaamse Instelling voor Technologisch Onderzoek Boeretang 00, B-00 Mol, Belgium Abstract Odour emissions coming from a landfill were quantified by using a reversed modelling technique in combination with a series of odour detection campaigns. During the detection campaigns an odour detection team traced the boundaries of the odour pollution area and at the same time the meteorological data needed for the atmospheric dispersion model IFDM were collected at the landfill site. Two different methods were used to compare the modelled odour distributions with the experimentally obtained data. The methods produced comparable results for the different campaigns, allowing an evaluation of the impact of the odour limiting measures currently applied to the landfill. Introduction Odour emissions can cause serious nuisance to people living in the vicinity of waste treatment facilities. In order to investigate and quantify the odour emission coming from a landfill for domestic waste located in Flanders, Belgium, a reversed modelling method was used in combination with a series of odour detection campaigns. During each of the campaigns, carried out in August and September 99, the boundaries of the odour pollution area were determined by an odour detection team, crossing the area near the landfill in various downwind directions during approximately hour. At the same time meteorological equipment installed on a 0 m mast measured the wind direction, wind speed and temperature. More details on the odour detection campaigns are given in section.

2 Transactions on Ecology and the Environment vol 8, 996 WIT Press, ISSN Air Pollution Monitoring, Simulation and Control The atmospheric dispersion model IFDM^ was used to calculate the odour pollution distribution due to the odour emissions coming from source locations at the landfill site. However, since the emission source strength was not known a priori, the distribution was presented in terms of so called "model units". This modelling technique is further described in section. Two methods were used to compare the modelled distributions with the experimental data obtained from the odour detection campaigns. In the first method the detected boundaries of the polluted area were marked and the modelled values located on these boundaries were used to quantify the emission source strength. The second method consisted in a direct mapping of the modelled distribution on the measured data using a screening technique. Through linear regression the odour threshold in terms of model units was obtained and from this result the emission source strength could be calculated. A detailed description of the two methods and their results can be found in section. In section the results of this reversed modelling technique as well as the method's uncertainties are discussed. Odour detection campaigns. Campaign settings In order to investigate the effect of the odour limiting measures currently applied to the landfill, a series of five odour detection campaigns were organised during August and September 99 (see Table for more details). The landfill for which the campaigns were organised is situated near a small town in Flanders and has been operational since 989. It is situated in an old clay pit, seized 700 x 00 m. The municipal waste is shredded and compacted before it is transported to the landfill site, where it is discharged in layers of to m thickness. The discharged waste is covered daily and a gas extraction and flaring system has been installed after a first demand for remedial actions in 99. The main source locations on the landfill site were associated with freshly tipped waste arriving at the mechanical treatment system and with the active depositing area. This is in agreement with emission estimations for domestic landfill sites reported by Frechen *'*.. Odour detection method Although it can provide very useful additional information^ a chemical analysis of individual odour components does generally not allow a quantification of the source strength of a mixture of odour components. Translation of the analysis results towards odour perception is another problem. In contrast to chemical analyses the human nose is a very accurate and perceptive sensor for the

3 Transactions on Ecology and the Environment vol 8, 996 WIT Press, ISSN 7- Air Pollution Monitoring, Simulation and Control 689 detection of odour generated by a mixture of components. It therefore forms the basis of many sensorial measurement methods. A commonly used sensorial method is olfactometry. In this method air is sampled and diluted with odourless air in an olfactometer in order to measure the odour concentration. The odour concentration is expressed in odour units per cubic meter of air (o.u./nf). One odour unit is defined as the dilution of the sample air in nf of odourless air which is (still) percepted by 0% of the panel members performing the perception test. This value is also known as the odour perception threshold. The number of times the air sample has to be diluted in order to reach the odour perception threshold is denoted as the odour concentration. For example: if a sample has to be diluted 00 times to reach the odour threshold, the odour concentration is said to be 00 o.u./nf. Unfortunately there is no standard for olfactometry tests or olfactometry equipment and odour thresholds can internationally differ by a few orders of magnitude. The choice of the panel members is important as well. Taking representative air samples to obtain a uniform picture of the odour emissions is another problem, especially for the situation of odour emissions coming from a landfill with multiple emission sources^. An alternative for olfactometry can be found in odour detection at the site. This is sometimes referred to as "sniffing campaigns" in literature*. A team of selected odour detection specialists is send out to determine the boundaries of the odour pollution area by crossing the area near the landfill in various downwind directions during approximately hour. The odour observations are indicated on a map ("x" if odour, "o" if no odour) and by combining the observations of the different team members a contour of the odour pollution area can be obtained. Although such methods are not standardised yet, guidelines were followed that have been published by various foreign authorities^'*. These guidelines impose the following conditions on the team members^ : the odour detection team should consist of at least 6 persons, the team members are selected on their individual odour threshold for n-butanol, conform NVN 80 *, the observation actions have to be co-ordinated by one person (who is not involved in the detection activities), * the team members are not allowed to discuss their findings during the observations. The selected odour detection specialists should meet these four objectives in order to be allowed to participate in the detection campaigns. At VITO the panel members were tested on their ability to make a distinction between various diluted ethanol solutions and diluted n-butanol solutions as described in the guideline * *.

4 Transactions on Ecology and the Environment vol 8, 996 WIT Press, ISSN Air Pollution Monitoring, Simulation and Control Other conditions imposed on the odour detection method are: the detection frequency should be minutes, the odour should be recognisable from background odour, the domain near the odour emission source should be accessible in the direction perpendicular and along the plume axis. The campaigns were carried out in a domain with typical dimensions of. km x. km. This domain was always located downwind of the landfill and was found to be sufficient to include all the odour observations.. Meteorological observations During the odour detection campaigns meteorological observations were made by the meteorological equipment installed on the landfill site. Wind direction (d [deg]) and wind speed (v [m/s]) were measured at 0 m height using an anemometer. A thermistor was used to obtain the temperature (T [ C]). The meteorological data were stored every seconds by a data logger. After processing the data they were offered to the dispersion model as minutely averaged values. Table gives an overview of the meteorological situations per campaign. The values in the last three columns in Table are averaged over the campaign duration. Table : Meteorological characteristics for the odour detection campaigns. NO, Date I. % Timey.: -ii, 9/08/9 /08/9 /08/9 /09/9 /09/9 0h0-hOO 9h0-0h0 hoo-hoo 9h0-0h0 0h-h0 E E E E E ;>:Vfm/sj>,o,..,,;, :#w MT; " j, > * \ V^i N V.? Except for wind speed and wind direction the atmospheric dispersion model needs a stability class parameter defined by Bultynck and MaletV This parameter is expressed as Ei with i varying from to 6. Its value is based on the temperature difference at 8 m and m and on the wind speed at 69 m. E; is a measure for the atmospheric stability, ranging from El (very stable), via E (neutral) to E6 (very unstable). It determines to a large extend the shape of the plume as will be shown in section.. For each campaign the Bultynck-Malet stability class parameter was obtained from a meteo-tower at 0 km from the landfill. The values obtained from this tower are given in the fourth column of Table.

5 Transactions on Ecology and the Environment vol 8, 996 WIT Press, ISSN 7- Air Pollution Monitoring, Simulation and Control 69 Reversed odour emission modelling. Calculation of trajectories In order to get an idea of the area affected by the odour emissions during the detection campaigns, air parcels (also called puffs) were followed from the emission sources on the landfill site towards the edges of the. km x. km domain. The air parcels were emitted at the active depositing area and from there transported by means of the minutely averaged values of wind speed and wind direction as obtained through the local meteorological observations. The calculated path which is followed by each puff is defined as a trajectory. b) campaign no. Figure : Calculated trajectories for al campaign no. and bl campaign no. Figure shows the trajectories for two different meteorological situations. In Figure la the trajectories were calculated for a neutral situation observed during campaign number, whereas in Figure Ib the trajectories are given for a slightly unstable situation observed during campaign number.. Atmospheric dispersion modelling using IFDM The bi-gaussian atmospheric dispersion model IFDM? was used to compute the emission source term. This model is used as a regulatory model in Flanders. It is used for impact assessment studies on a local scale (up to 0 km) and it can calculate air pollution due to non-reactive gases coming from both point and area sources.

6 Transactions on Ecology and the Environment vol 8, 996 WIT Press, ISSN 7-69 Air Pollution Monitoring, Simulation and Control In this model application the landfill is represented by 8 source locations divided equally over the mechanical treatment system where the freshly tipped waste arrives and the active depositing area. The location of the active depositing area varied from campaign to campaign and had to be defined on site. For each source an emission source strength of "model units" per second was assumed. So the total emission source strength was model units per second. The source height was assumed to be m. The domain of. km x. km was discretised by 6 receptor points in each direction (x and y), giving a spatial solution of 00 m. As meteorological input data the model used the minutely averaged values for wind speed and wind direction together with the stability class parameter E; valid for the duration of the campaign, as described in section.. The model generates an immission distribution with concentrations presented as model units per nf. Figure gives this immission distribution in model units per nf for campaign number.. km 6 U S* i S. 6 JL ±, Z 0 7 A. a n 0 7. Z? - 8 Z % js. % % T i ' Figure : Odour immission distribution calculated bv IFDM for campaign no.

7 Transactions on Ecology and the Environment vol 8, 996 WIT Press, ISSN 7- Air Pollution Monitoring, Simulation and Control 69 Determination of the odour emissions. Methods combining reversed modelling and odour detection results With the odour detection map obtained by the odour detection team on one hand and the calculated odour immission distribution in model units (Figure ) on the other hand, the aim of the reversed modelling method is to find the corresponding relation between the odour detection threshold (= o.u./irp) and the immitted model units per nf calculated by the dispersion model. This relation was established by applying two different methods. The first method was named the plume edging method, the second was called the plume screening method... Plume edging method This method assumes that the odour detection threshold is situated at the edge of the plume, right between the locations where odour was detected (x) and locations where no odour was detected (o). The boundaries of the polluted area were marked in the immission distribution grid by ticking the model units on both sides of the plume edge (Figure ). The average value gives the number of model units corresponding to the odour detection threshold at this location. By summarising these values and averaging again over the number of locations where the plume edge was captured, an over-all value for the odour detection threshold is obtained... Plume screening method The second method consisted in a direct mapping of the modelled immission distribution on the odour detection data. A screen with a mesh size of 00 m times 00 m (equivalent to the resolution of the model grid) was laid on top of the map with the odour detection data. Each squared mesh cell was given a number between 0 and depending on the amount of crosses (x) and zeros (o) observed in this mesh cell. If only crosses were observed a value is given to this particular cell. If only zeros are observed a value 0 is given to the mesh cell. A linear interpolation is used for the values between 0 and, for example: two crosses and three zeros in one cell results in a value 0.. Mesh cells that had no odour observations (x or o) included were not taken into account. In the next step the odour values between 0 and as obtained per mesh cell were plotted versus the model units found in the same mesh cells of the calculated immission distribution. Figure shows the mesh cell values between 0 and in function of the model units for campaign number. In the cases where more than one mesh cell value was found per model unit number, these values were averaged. The odour threshold in terms of model units was obtained at the intersection of the linear regression line through the points with the dotted line indicating an odour cell value of 0. (see Figure ). Note that an odour cell value of 0.

8 Transactions on Ecology and the Environment vol 8, 996 WIT Press, ISSN 7-69 Air Pollution Monitoring, Simulation and Control. 0.8 CD 0.6 ^ o ' Model units Figure ; Odour cell values versus model units for campaign no. means that 0% crosses and 0% zeros were found in this cell, which is equivalent to the odour detection threshold. The number of model units found in this way are then used to calculate the emission source strength.. Results For each campaign the results for the emission source terms are summarised in Table. Table : Results for the emission source strength calculations.ho: ' -M - M ~^>,~CorrJ > > *& o,y Js ^[aujifl In the columns Ml and M the number of model units corresponding to the odour detection threshold ( o.u./nf) is given for the first and the second method respectively. The fourth column Corr. gives the correlation coefficient obtained for the linear regression used in the second method. The last two columns provide the emission source strength SI and S [o.u./s] for the two

9 Transactions on Ecology and the Environment vol 8, 996 WIT Press, ISSN 7- Air Pollution Monitoring, Simulation and Control 69 methods, calculated from Ml and M respectively with the emission strength in model units known as model units/s. Discussion The value for the emission source strength was found to be 70 o.u./s (± %) for thefirstmethod and 9 o.u./s (± 6%) for the second method. The difference between the two results was found to be in the order of 0%. Although the plume edge method seems to provide the results with more certainty, it is clear that this method assumes a sharp plume edge. This is not always the case. There can be a zone in which the transition from odour to noodour is not very clear. The plume screening method tries to take into account this zone by allowing mesh cells to have values between 0 (no-odour) and (odour). In order to find out the sensitivity of the model results towards the location and the distribution of the sources, a computation for one single source emitting model units per second was performed. This source was located at the active depositing area. The distributed model units near the plume edge were found to vary with one unit or less. This means that this sensitivity run results in an error of max. 8% in the computation of the emission source strength. The uncertainty on the source location therefore remains a problem, although arguments can be found in literature *' * to support the choice of the source locations near the mechanical treatment system and in the vicinity of the active depositing area Other uncertainties in determining the emission source strength can be found in the fact that it is assumed that the odour is emitted at a constant rate throughout the campaigns. Although temperature variations were small during the campaigns (see Table ), it might be argued that temperature as well as humidity might have an influence on the emission source strength. This has no impact on the method itself, but it may hamper the evaluation of the remedial actions that were taken to reduce odour nuisance. 6 Conclusions A reversed modelling method was used in combination with odour detection campaigns to determine the odour emission from a landfill for domestic waste. Two different methods were used to calculate the emission source strength. The methods produced comparable results (within 0%) for the different campaigns. The inaccuracy of the methods was found to be to 6 %. The uncertainty in the location of the source(s) was estimated to be 8%.

10 Transactions on Ecology and the Environment vol 8, 996 WIT Press, ISSN Air Pollution Monitoring, Simulation and Control 7 References. Bultynck, H., Malet, L Evaluation of atmospheric dilution factors for effluents diffused from an elevated continuous point source, Tellus, 97,, -7.. Cosemans, G, Kretzschmar, J., Maes, G. The Belgian Immission Frequency Distribution Model IFDM. Proceedings of the DCAR Workshop on objectives for next generation of practical short-range atmospheric dispersion models, Ris0, Denmark, pp 9-0, 99.. De Bree, F.B.H., Harssema, H. Application of sniffing teams within odour pollution research, Proceedings of the 8th World Clean Air Congress 988, The Hague, The Netherlands, - Sep Frechen, F.-B. A new model for estimation of odour emissions from landfill and composting facilities, Proceedings Sardinia 9, Fifth International Landfill Symposium, S. Magherita di Pula, Gagliari, Italy, - 6 October 99, pp 8-88, CISA, Cagliari, 99.. Frechen, F.-B. Air pollution caused by odours - basics, measurement techniques and emission inventory at waste and wastewater treatment facilities, in Pollution Control and Monitoring, Air Pollution II, ed JM. Baldasano, et al., Vol., pp. 86-9, Computational Mechanics Publications, Graafland, T.F., Anzion, C.J.M., Dragt, A.J Bruikbaarheid snuffelploegmetingen bij stankoverlast, VROM Publikatiereeks Lucht 66, Staatsdrukkerij, 's Gravenhage, Griffin, L.R., Rutherford, T.L. Comparison of air dispersion modeling results with ambient air sampling data: a case study at Tacoma landfill, a national priorities list site, Environmental Progress, 99, Vol.,, NVN 80, Luchtkwaliteit. Sensorische geurmetingen met een olfactometer, ledruk, Paduch, M. Festlegung und Beurteilung von Geruchsimmissionen (Geruchsimmissions-Richtlinie), Staub - Reinhaltimg der Luft, 99,, Smith, W, Valk, C.J., Anzion, CJ.M. De meetnauwkeurigheid van snuffelploegen, Witteveen+Bos Raadgevende ingenieurs b.v., 99.