Following the odour pathway

The use of dispersion models to predict the odour concentrations emanating from WwTWs has become commonplace. The resulting odour contour plots are often used in support of planning applications to show nuisance will be minimal or non-existent. Dispersion models have undergone a great deal of development and the latest models are much improved in terms of accuracy, resolution and their treatment of complex topography.

Unfortunately, dispersion is only one aspect of odour nuisance. It is important to consider the whole annoyance pathway from source to receptor (Figure 1). Each element

in the pathway is equally important in determining the potential for annoyance.

An over-emphasis on dispersion modelling, where the other elements of the pathway are treated simplistically,

can give an inaccurate prediction, even though the graphically impressive odour

contour plot gives a sense of accuracy and precision.

Modelling the entire pathway is difficult. One of the main problems is the difficulty in measuring odours. We can either measure odorant (odorous molecules) concentration using analytical means or we can measure the effect of the odorants on the sense of smell using dilution methods (olfactometry). The relationship between odorant concentration and odours as perceived is not straightforward and is complicated by the fact that sewage emissions may contain hundreds of different odorants.

A common alternative is to measure the dominant odorant analytically and to use this as an indicator of the overall odour strength. Hydrogen

sulphide (H2S) is often used for this purpose because it is one of the most common odorants associated with sewage and is relatively easily measured in both gas and

liquid phase. Odour perception is also a difficult area.

Sensitivity and attitude to odours vary significantly amongst individuals, although on aggregate it is possible to produce dose-response curves for odour annoyance.

Typically, odour exposure (the dose) is estimated using dispersion modelling, and the degree of annoyance felt by the population (the response) is assessed using questionnaire-type approaches.

Dose-response study results have been used to determine odour limits, for example, the maximum concentration allowed at either the site boundary or the nearest receptor. These are usually specified on a concentration and percentile basis and are also influenced by the averaging period used by the dispersion model. Odour limits are the subject of much debate. Because of concerns over the over-emphasis on dispersion modelling and the lack of attention given to inputs to dispersion models, recent work at Cranfield University has concentrated on the development of a linked formation-emission model for H2S.

We have called this model the Sewage Treatment Emission Calculator for H2S (STEnCH). The model employs a liquid-phase sulphide model developed from a sewer model that was developed by Hvitved-Jacobsen et al (2000) and uses mass-transfer models for wind-blown surfaces, weirs, drop structures, channels, aeration tanks and trickling filters.

This combination of liquid-phase and mass-transfer models allows a variety of sewage treatment processes to be modelled. Further details

can be found in Gostelow

et al (2001, 2003).

The STEnCH model allows the influence of process, flow, quality or meteorological variables to be examined. Because the models are dynamic, the impact of these variables on temporal emission rate variations can be examined.

To illustrate the potential for emission rate variability, two hypothetical cases have been studied. The first considers the emission rate variation resulting from wind effects on primary sedimentation tanks and the second considers the effect of wastewater flow and quality variations on emissions from primary sedimentation tanks fed by a long sewer.

For sedimentation tanks, surface emissions are largely driven by wind-induced turbulence. It is very uncommon for this to be taken account of in odour modelling exercises.

To illustrate the importance of wind-influenced emission rates, two cases have been considered. The first utilises a fixed emission rate derived for the average wind speed of the one-year meteorological data set used for dispersion modelling. The average wind speed is 3.96ms-1, which, using the STEnCH emission model, gives a surface emission rate of 2.42 x 10-7 g s-1 m-2. The second case allows emission rates to vary with wind speed. Figure 2 shows the distribution of the hourly averaged wind speed and the corresponding modelled surface emission rate distribution for the one-year data set.

This shows the wind and emission rate distributions to be significantly negatively skewed with lower values being more common than higher values. Dispersion modelling results using the USEPA ISCST3 model are shown in Figure 3.

There is a large difference between odour footprints for the two cases. For the fixed emission rate case the 1ppb contour extends approximately twice as far as for the variable emission rate case. The reason for this is the skewed nature of the emission distribution.

This means lower emission rates are more common than higher ones. Also, when higher emission rates do occur, this is when wind speeds are higher and dispersion improved. This is why the use of a simple average fixed emission rate can result in an over-estimate of odour impact.

Because of variations in the flow and quality of wastewater entering a WwTW, there is the potential for significant variations in influent odorant concentrations and subsequent emissions in the downstream treatment processes.

To investigate these effects, the STEnCH sewer and sedimentation tank models have been used, considering a set of primary sedimentation tanks at the end of a 10km sewer.

The simulation has been engineered such that the sewer flows full at flow rates above the average daily flow. During these episodes, conditions will become anaerobic in the sewer and sulphide forms. Conversely, for flow rates below the daily average, the sewer flows part full and dissolved oxygen can again be supplied.

During these periods, conditions are aerobic and any sulphide formed can be oxidised. Figure 4 shows the emission rate variations predicted by STEnCH for the sewer-sedimentation tank system. Emission peaks are observed relating to the high flow conditions, because the sewer flows full during this time and sulphide is formed, and also because emissions from the sedimentation tank weirs are enhanced by high flows. The large variations in emission rates resulting from flow and quality variations has implications for odour measurements. For example, if spot emission rate measurements were being used as inputs to dispersion models, the time of the measurement would be crucial.

Figure 5 shows the different ISCST3-predicted odour footprints that would result from emission measurements taken at 9am and 4pm.

The resulting odour footprints are very different. The outcome of a dispersion modelling exercise depends on the quality of the input data used.

Unfortunately, this is often overlooked, with emission rates being poorly estimated and usually regarded as constant. This can lead to inaccurate predictions of odour footprints due to the lack of consideration of emission variations caused by meteorological, flow, quality or process variations. There is currently considerable debate regarding suitable limiting concentrations and percentiles to prevent odour annoyance.

Unfortunately, these arguments may be flawed because only the effect of the meteorological variables is included

in these percentiles.

Where there is sig-

nificant variation in emission rates, as is the case for WwTW, the variations

can radically alter the

odour footprint. This is

why it is essential for emission rate variations to

be included in odour

modelling exercises, and

integrated odour modelling approaches developed.