Login 
This article describes a new approach, which aims to create bespoke modelling methods that are more refined for site-specific predictions.
Introduction
The following sections will introduce the background of dust propagation, software modelling, risk predictions, and objectives.
Background
‘Dust’ is defined in BS 6069 Part 2 (BSI, 1994) as particulate matter up to 75µm in diameter. One expression for the coarser fraction of this is Total Suspended Particulates (TSP). Other expressions, such as ‘nuisance dust’ and ‘fugitive dust’ are used, and these terms are also generally taken to mean ‘visible particulate matter’. Unfortunately, such expressions are poorly defined and as a consequence there is no satisfactory measurement method for such material.
The effects of this coarser fraction can be both chronic and acute, from the long term soiling of surfaces to the immediacy of visible dust plumes. The finer size fractions of PM10 and PM2.5 (dusts with an aerodynamic diameter less than 10 or 2.5µm) are measured much more accurately and are closely monitored because of their association with impacts on health, through inhalation. Concentrations of PM10 and PM2.5 are therefore used as indicators of local air quality through the National Air Quality Strategy (NAQS) and associated Air Quality Objectives (AQO) (Defra, 2007).
Fugitive dust movements from industrial sites have the potential to cause problems in surrounding areas, often through its accumulation over time. Emissions of such material can be monitored through dust sampling, in which dust monitors are positioned to sample dust both in flux and in settlement. Ideally, these devices are deployed as an array to monitor dust within, on and beyond site boundaries, and to evaluate dust levels from a range of sources and in the directions of off-site receptors. In certain circumstances, however, this may not always be feasible.
Sites might be undergoing expansion or change, access might not be permitted or the project might simply be in the planning stage. Accurate dust predictions, however, might still be required by regulatory bodies or operators for planning applications. Consequently, meaningful prediction of the potential risks that are posed to nearby receptors might be required.
Predicting the levels of dust emitted from individual sources, and more importantly whole sites, can be problematic for a multitude of reasons including but not limited to the local topography, weather conditions, the nature of dust-generating operations, distance and directions between source and receptor, levels of screening, and even the local demography due to several potential sources. Current methods to resolve the problem may be split into two broad approaches: software modelling and risk prediction based on professional judgement.
Software modelling
Software models designed for use in air quality investigations may be used for nuisance dust predictions. Those normally employed are Gaussian dispersion models which require complex source emission data to be accurate. They are typically used to investigate air quality, often in urban areas, using plume dispersion calculations for gases such as NOx (nitrogen oxides) and SO2 (sulphur dioxide) and finer dusts such as PM10 or PM2.5.
For such studies, emission inventories for point sources such as stacks, or linear sources such as roads, are readily available in the UK and abroad. Problems occur, however, when pollutants such as visible, ‘nuisance’ dust are modelled; such pollutants are generally associated with area sources and reliable emission inventories for such material are very limited. Available data include basic estimates such as those by the European Environment Agency (EEA); for example, the EEA indicates that the storage of minerals, without dust suppression, would give rise to 8.2 tonnes of total solid particulates (TSP) per hectare per year. Other such calculations at mineral sites, available both from the EEA and the US Environmental Protection Agency (EPA) are based on emission contributions per unit weight of material transferred.
Many estimates do not, therefore, take into account any site or substrate information, let alone any temporal fluctuations in weather conditions. Even those that do have been shown to be inadequate compared to corresponding observations (Venkatram, 2000).
Risk prediction
The other common approach uses tabulated risk values based on professional experience to assess likely dust nuisance based on factors such as the source to receptor distance, the number of receptors and their sensitivity, the site size and type and typical weather conditions in the area (e.g. IAQM, 2014). Using a combination of these, calculations can be made on the likely magnitude of impacts for fugitive and windblown dusts. Such procedures are useful for analysing a site’s impact on a surrounding area, but are not able to provide predictions of dust levels or the likely conditions for significant dust events.
Objectives
This article describes a new approach that aims to create bespoke modelling methods that are more refined for site-specific predictions. A case study is presented where a new approach to software modelling is shown using original dust monitoring data. An alternative assessment using data from the literature is shown as a comparison with site-specific modelling.
Methodology
Equipment
This study used sticky pad dust monitors to assess nuisance dust movements. Passive sticky pad dust monitors can be used to collect dust both in flux on directional sticky pads mounted on cylinders for 360° capture, and in deposition on horizontal mounts. Samples are routinely taken over seven or 14 day intervals before being scanned and analysed using custom software.
Dust coverage (Absolute Area Coverage – AAC%) and dust soiling (Effective Area Coverage – EAC%) are then reported for each 15° segment (or the average for dust settlement samples) that indicates the direction in which the dust was travelling at the moment of capture and the potential risk of nuisance. The adhesive allows the dust to be removed from the sticky pad for filtering and reporting in total mass, as well as for further characterisation such as elemental or mineral analyses or particle size grading. This enables dust to be ‘fingerprinted’ to separate dusts from their respective sources (Fowler et al , 2013).
Case study site
A central England mineral site was used for a trial study into both assessing bespoke empirical models and in the first steps of developing site specific emission factors for use in existing modelling software. The site chosen had two years of available dust monitoring data with eight directional and depositional sticky pad dust monitors located around the site. Two were located on the site boundary and a row of five sampling stations ran from south west to north east; the direction of the prevailing wind. An electronic weather station was also located on the western boundary of the site. The site management team assisted in supplying supporting data, including site movements and activities.
The first steps taken were to examine the accuracy of existing emission factors from the literature by comparison with collected samples. The emission factors used were taken from guidance from the EEA and the EPA and were as follows:
1 The EEA‘s 2013 Guidebook for ‘Storage, handling and transport of mineral product’ (2.A.5.c) states that 16.4 tonnes of TSP will be emitted per uncontrolled hectare per year, with 95% confidence intervals of 8.2 tonnes and 32.8 tonnes. Controlled emissions will be reduced by a factor of 10. PM10 emissions are assumed to account for 50% of the total.
2 The US EPA’s AP42 guidance, section 13.2.4 for ‘Aggregate Handling and Storage Piles’, provides an equation for emissions from stockpile transfers that uses wind speed and moisture content to calculate emissions of dust in kg per tonne of material transferred.
Using data from the on-site weather station and the quarry site movements, emission rates were calculated on an hourly basis through a six month period when dust sampling had been undertaken. These then provided an input parameter for a commercially available software package, which is widely used in the UK for environmental pollution modelling.
The following additional data were required:
• Source, monitoring location and receptor coordinates • Weather data including wind speed and angle, cloud cover, rainfall and temperature • Terrain data for the site and the surrounding area
The site was treated as one large area source, with emission rates calculated as ‘constant’ from mineral storage piles and the quarry floor and haul roads, and ‘variable’ for emissions from site movements and mineral transfers. Modelling outputs were both as a grid for contour mapping, and individual points for comparison with relevant dust samplers, with hourly averages for the whole six month period. Weekly concentrations were calculated for comparison with the dust monitoring data.
Dust monitoring units
In order to compare dust sampling results with the outputs of the modelling software, which were given in dust mass per unit volume (mg m -3 ), it was necessary to calculate the mass per unit volume. The dust sampling collection methods outlined provided results for dust soiling of the sticky pads and, after additional laboratory analysis, in dust mass per unit area of the sticky pad (mg m -2 ).
An additional dust sampler with a known efficiency was therefore located next to a sticky pad dust monitor to assess the efficiency variations. A linear regression model was prepared to evaluate the efficiency of the sticky pad sampler by comparing the two sets of results in addition to wind speed data. This enabled sticky pad dust sample data to be converted into mg m -3 .
Results
The first set of results showed that the overall level of dust predicted using the published emissions factor inventories were of the right order (93µg m -3 predicted compared with 54µg m -3 measured) but did not pick up the temporal variations that occur from week to week (Figure 1). Most importantly, the period with the highest dust was predicted using the published emission factors, but the level of dust modelled was far lower than that measured on site by a factor of 3 (148µg m -3 predicted compared with 436µg m -3 measured). The second highest monitored dust incident was not predicted at all.
In the context of these findings it was inferred that a method for back calculation might be possible to establish effective emission factors and thence improve the accuracy of the models.
New methodology
In order to establish effective emission factors based on site dust monitoring data, the previous model results were adjusted with reference to the dust levels obtained from a boundary monitor at the site. There was a linear relationship between source emissions from the site and concentrations at the boundary, so a simple adjustment was made to ‘back calculate’ the emission rate at the source over the whole six month period. Emissions were adjusted for each half hourly period and the model was subsequently run again with the revised emission rates and tested against different sampling locations.
Results
Figure 2 shows that at two different dust monitoring points the new predictions matched the sample data very closely (r 2 > 0.89). The average concentration throughout the period also compared well at both locations (25 µg m -3 actual at monitor 1 compared with 21µg m -3 predicted, 12 µg m -3 actual at monitor 2 compared with 10 µg m -3 predicted). The results therefore showed that the model accurately predicted the dispersion of dust for the prevailing site and weather conditions, and that once the emission rates were established useful predictions could be made.
Discussion
Nuisance dust movements can cause significant problems within and around some industrial sites; considerable expense can be incurred in installing and using dust suppression equipment that may only be required occasionally. Settling local disputes may also be difficult and awkward if further planning permissions are needed. It is important, therefore, that any predictive work is accurate and well-founded to enable sound judgement of potential or existing problems. The initial results using published emission factors show that there is room for improvement in nuisance dust predictions.
Many published emission rates do not take into account site conditions or any seasonal variations and are given as flat rates, whereas ground and weather conditions may vary greatly. Professional experience, however, shows that dust movements are largely dependent on weather conditions that in turn may influence ground conditions. The key to using models to their full capacity is to accurately measure or back calculate emission rates.
The method described above demonstrates a new approach to modelling nuisance dust movements at industrial sites. What has been presented is the back-calculation of dust from a site as a whole. The apportionment of dust to specific sources can be further aided by elemental modelling of specific dust sources and their contribution to dust at the site boundary or receptor. It is therefore possible to further understand the extent to which a specific source contributes to the cumulative dust at, say the receptor or site boundary. Often on large construction sites and quarries it is apparent that there may be several significant sources of dust.
Some indication of which source might be most significant may come from a visual analysis of the dust movement patterns. By means such as this a database can be built up of specific emission factors for individual site activities, which can be a powerful tool in predicting fugitive dust movements at sites with no dust monitoring.
Conclusions
The collection of dust monitoring and weather data is the key to calculating site-specific dust emission rates. This should have regard to source-pathway-receptor objectives and to the prevailing wind directions.
From an established dust emission rate, data from a single and appropriately located dust monitor may be used to predict dust dispersion to other locations and in other directions with the proviso that weather conditions are broadly similar to those obtained during the dust monitoring period.
When a comprehensive database of dust emission factors has been established this information could be used to predict dust dispersion at sites with no data or with no current dust generation. Such circumstances may arise when assessing new construction projects. This work will depend on a good understanding of the equipment and activities to be used.
This approach can also be used to assess the importance of different components of site dust where several sources of dust contribute to the dust that is collected at or beyond a site boundary. The ability to assess changes in the chemical and physical properties on the dispersion of dust compared with those properties at dust sources is currently under investigation.
Published: 14th Dec 2015 in AWE International
