Air quality assessments inform air quality management activities by providing an understanding of how pollutant sources, emission characteristics, topography and meteorological conditions contribute to local air quality.
Source Apportionment, Mobile Monitoring and Land Use Regression
1. Source Apportionment
Source apportionment techniques aim to estimate, or apportion, the contribution of different pollution sources to ambient concentrations within a given area.
Source apportionment is conducted by first understanding the particular make up of a mixture of air pollution, then linking these pollutants to specific sources.
Understanding the composition of a pollution mixture is an important step in determining the potential health impacts of exposure.
Types of Models
Models of varying complexity have been developed to conduct source apportionment. These models aim to attribute ambient pollutant concentrations at specific locations (receptors) to specific sources.
Common models include chemical-mass balance (CMB), principle component analysis (PCA) and a related technique called positive matrix factorization (PMF).
The key difference between these models is the requirement of prior knowledge/data; while CMB models rely on chemical source profiles of emission sources, as well as on the chemical composition of ambient air at receptor locations, PMF and PCA require only chemical composition information at receptors.
In order to conduct a CMB assessment for PM, filter samples are collected and undergo laboratory analysis to determine the composition of the collected samples. Additionally, the chemical profile of pollutants emitted from all major sources in the air shed must be specified.
The contribution of each source to the filter chemical make-up is then calculated by combining the sources linearly. The method is improved when sources have unique chemical tracers, making it easier to match the filter PM chemical composition to a specific source.
As very few pollutants are source-specific, only non-reactive chemical tracers may be used as indicators of specific pollutants.
For example, levoglucosan, a tracer for wood smoke, is often used in the analysis of PM to apportion the contribution of wood burning to a particulate sample.
Unlike CMB, PCA and PMF models can be used when the chemical composition of emissions from potential sources are unknown. PCA and PMF are very similar; both are statistical models that use multivariate receptor analysis to identify sources of a pollutant mixture.
Despite these similarities, PMF is thought to be superior to PCA for several reasons. In PMF, unlike for PCA, it is possible to account for missing data, values below the limit of detection and uncertainties in each of the data values, by assigning weights to the data values.
PMF is also more realistic since negative concentrations are excluded, unlike in PCA. Using both models, the chemical constituents of a sample are analyzed and the relationships between the constituents, expressed as a covariance matrix, are investigated.
When particular chemical species vary together, they are assigned to the same factor. The chemical make-up of each factor is then interpreted and identified with a specific source.
Information from a PMF model can be further refined with the use of meteorological data, including wind direction, to provide better information on the geographical location of the source.
For example, if a particular factor occurs when wind is from a specific direction, and the factor is chemically associated with a source in that direction, then the factor may be attributed to that source.
2. Mobile Monitoring
Mobile monitoring uses a mobile platform, typically a vehicle, to collect pollutant measurements across an area of interest.
This type of monitoring is useful to:
(1) Provide insight about areas that are not well represented by fixed-site monitoring stations;
(2) Capture small-scale spatial variability of pollutants;
(3) Identify localized pollutant hot spots, particularly for emissions that vary in concentration over small spatial scales, such as residential wood burning and traffic;
(4) Provide data for model development or validation. Mobile monitoring has the capability of being rapidly deployed, therefore, can also be used in emergency situations, such as characterizing the spatial distribution of a chemical plume resulting from an accidental release or smoke from forest fires.
For these reasons, mobile monitoring provides detailed information beyond what can be typically characterized by traditional fixed-site monitoring networks, so can be used to improve exposure estimates and inform air quality management decisions.
Mobile monitoring is typically conducted by equipping a vehicle with air pollutant monitors. The use of a geographical positioning system (GPS) allows precise locations to be assigned to air pollution measurements.
Read Also : Tools for Air, Water and Soil Analysis
There are two sampling methods that can be used to conduct mobile monitoring:
(1) Measurements can be collected while the vehicle is in motion or
(2) Can be stationed for periods of time at designated locations.
Generally, the purpose of collecting measurements while the vehicle is in motion is to gather a high density of measurements over an area of interest. Continuous monitors are suitable for this approach and are typically used to collect real-time measurements at high frequencies (less than 1 minute).
Several types of pollutants such as PM4 and air toxics have been measured, using this sampling method. Some measurements, such as PM, do not provide source-specific information, making it difficult to attribute specific sources to the mobile measurements.
However, different techniques, such as choosing an appropriate sampling period or instrument selection, can help to identify or isolate the sources of interest. For example, to characterize PM2.5 generated from residential wood burning, mobile monitoring can be conducted during cold, calm winter evenings.
During these conditions, wood burning activity is expected to be relatively high while the relative contribution of traffic to ambient PM2.5 is expected to be lower.
Selecting an instrument, such as a multi-wavelength aethalometer or multi-wavelength nephelometer, instruments that measure light attenuation and scattering of a sample (respectively) at two or more wavelengths, can help to distinguish some particle sources, such as diesel exhaust or wood smoke.
Supplementary sampling at fixed-sites can also help to characterize the chemical contents of PM2.5 within the region of interest.
In cases where sampling is conducted while the vehicle is stationary, the vehicle essentially serves as a temporary monitoring station. The vehicle is stationed at designated sampling sites in an area of interest for longer sampling periods (from hours to days).
Sampling can be conducted for pollutants such as: PM, NO2, SO2, ozone, VOCs, PAHs, other air toxics, as well as for meteorological conditions with continuous and non- continuous monitors.
This approach is useful for obtaining ambient air quality information that would otherwise not be available through existing fixed-site monitoring networks.
3. Land Use Regression
Land use regression (LUR) is a modelling approach that can be used to describe the distribution of air pollution within urban and suburban areas. It was first developed by public health researchers in the mid- 1990s to examine neighborhood-scale variability in long-term concentrations of urban air pollutants.
At the time, there was new evidence to suggest that increasing exposure to city-wide air pollution had a negative impact on important indicators of public health.
LUR was developed to support epidemiologic studies investigating the public health effects of air pollution due to ambient air pollution variability within a single city. More recently, LUR has gained attention in the air quality management and urban planning communities.
Although LUR is typically used to model pollution related to vehicle traffic, the method has also been applied to sources like residential wood smoke and marine traffic. Regardless of the source under consideration, the premise of any LUR model is that the pollutant concentration at a specific location is a function of the physical characteristics of that location and its surroundings.
For example, LUR assumes that the nitrogen oxide (NOX) concentrations around a house may be associated with the volume of traffic around that house.
Likewise, the concentration of wood smoke-related particulate matter (PM) around that house may be related to the density of houses in the neighborhood with wood-burning appliances. The concept is easy to understand and the method is generally straightforward to apply.
There is no standard way of conducting LUR, but detailed descriptions of different approaches can be found in the scientific literature. The first step is always to measure a pollutant at multiple locations around an area.
These locations are generally fixed, but mobile monitoring has been used in some cases. Under ideal circumstances, the sites are specifically selected to optimize the spatial variability in pollutant concentrations.
Physical and geographic characteristics that might be associated with those concentrations are measured around each site, using a Geographic Information System (GIS). These potentially- predictive variables typically describe site location, including land use, population density, and traffic patterns.
Once sampling is complete and the potentially-predictive variables are generated, multiple linear regression is used to determine the association between measured concentrations and the most predictive variables.
The resulting equation can be used to estimate pollutant concentrations wherever all of the predictors can be measured; concentration maps with high spatial resolution can be generated by rendering the regression model in the GIS.
Studies to date have used a variety of methods to choose sampling locations, from convenience sampling (i.e., using a pre-established air monitoring network) to sophisticated location-allocation models that optimize the estimated variability in measurements while maximizing the distance between samplers.
While there is little evidence to support using any single method, LUR is most informative when models are built on data that reflect the full within-area variability of the pollutant in question. Likewise, there are no definitive guidelines on the number of sites to sample, but to capture the necessary variability, a practical minimum of 40 has been.
Finally, the sampling period should be chosen to suit the specific objectives of the study. For example, if using LUR to predict the long-term average of a pollutant that follows distinct seasonal trends (i.e., NOX), it is advisable to sample during periods that are known a priori to approximate the annual mean.
For the other side of the regression equation, it is important to consider which data will be used to generate the potentially-predictive set of variables.
Although the availability of geographic data depends upon local circumstances, most LUR studies on traffic-related pollution have used variables that quantify traffic intensity (sometimes specified by vehicle class), road classification density, distances to certain road types, population/building density, areas of land use classifications, and topography.
Some studies have attempted to improve model fit by including a wider range of data from other sources, such as meteorological models and remote sensing platforms.
In general, LUR can accommodate any spatial dataset that may help to describe the within-area variability of pollutant concentrations.
In summary, the different air quality assessment tools we have discussed can be used to identify particular sources that are important contributors to local air pollution, thus, enabling them to be targeted for emissions reduction strategies.
Information about important polluters can inform decisions on emissions permitting and industrial siting in a region. A better understanding of sources and of pollution composition can also better inform the assessment of health impacts once the exposure is better characterized.
Read Also : Palm Fruit: Health Benefits, Facts and Recipes