How is particulate matter measured

Of the myriad complex interrelated potential explanatory variables, only a small number have been used in the modelling of PM 10 concentrations.

One key factor commonly used to explain and evaluate trends in PM 10 data is the impact of meteorological conditions. The thickness of the ABL can vary from to m and extends from the ground to the point where cumulus clouds form.

In the ABL wind, temperature and moisture fluctuate rapidly, and turbulence causes vertical and horizontal mixing. Suspended in the ABL, particles may undergo physical and chemical transformations triggered by factors such as the amount of water vapour, the air temperature, the intensity of solar radiation and the presence or absence of other atmospheric reactants.

It is these physical processes, which help to explain why meteorological variables have such an influence on PM 10 concentrations. Having accurate and complete input data is critical to the success of any PM 10 prediction model. As a result, most models make use of data that are readily recorded using weather station sensors. In cases where data are incomplete, the instance is often removed rather than imputed because of errors which may be introduced by estimation processes.

The outputs of numerical weather forecast models can also be used as input variables in PM 10 models. However, this is not common because of the uncertainties such variables introduce to PM 10 predictions [ 14 , 15 ].

Wind speed and temperature are the meteorological explanatory variables most frequently used in PM 10 prediction models Table 1. Wind variables have been found to be useful proxies for physical transportation factors; wind is critical to the horizontal dispersion of PM 10 in the ABL. Wind direction controls the path that the PM 10 will follow, while wind speed determines the distance it is carried and the degree to which PM 10 is diluted due to plume stretching.

The effect of wind speed and direction on PM 10 varies with the geographical characteristics of a location. Low wind speed can be associated with high PM 10 [ 16 , 17 ]; this is common in hilly or mountainous regions. Conversely, in coastal or desert regions, high wind speeds result in high PM 10 concentrations due to salt or dust suspension. In Europe, PM 10 concentrations are significantly influenced by long-range transport contributions, which are independent of local emissions, so both wind direction and speed have a significant impact [ 18 ].

In Invercargill, New Zealand, where there are no close neighbours and thus little long-range transboundary PM 10 , wind speed explains most of the variability in PM 10 concentrations [ 19 ]. Cold temperatures increase the likelihood of an inversion layer forming in many locations.

An inversion prevents the upward movement of air from the layers below and traps PM 10 near the ground. As a result, cold temperatures tend to coincide with high concentrations of PM However, in some locations days with high temperatures, no clouds and stable atmospheric conditions result in high PM 10 [ 17 ].

In other locations when the difference between daily maximum and minimum temperatures is large and the height of the ABL mixing layer is low, high PM 10 concentrations are observed [ 20 ].

PM 10 levels can be reduced by rain, snow, fog and ice. Rain scavenging, a phenomenon in which below-cloud particles are captured and removed from the atmosphere by raindrops, is considered to be one of the major factors controlling the removal of PM 10 from the air.

The degree to which PM 10 is removed is dependent on rainfall duration and intensity [ 21 ]. While rainfall is a primary factor in PM 10 concentrations, it has not been used widely in models. This is in part due to the fact that in some countries, there is no rain for long periods of time or little rainfall in summer. The lack of rain data means that it is not often included in PM 10 models [ 14 ]. Relative humidity has been used more frequently in models than rainfall.

The relationship between PM 10 concentration and relative humidity also depends on other meteorological conditions. For example, if humidity is high and there is also intense rainfall such as during a monsoon season , then humidity has a negative correlation with PM 10 due to rain scavenging.

If high humidity is not accompanied by rainfall but is accompanied by high temperatures, humidity has been found to contribute to higher PM 10 concentrations. High solar radiation has also been shown to result in lower PM When solar radiation is high, the surface of the earth is warmer; as a consequence the exchange of heat in the air results in turbulent eddies that disperse suspended particles [ 23 ]. Autocorrelation is a basic structural feature of the meteorological variables used in PM 10 models.

When a numeric time series correlates with its own past and future values, this is known as autocorrelation or lagged correlation. A positive autocorrelation indicates persistence and a tendency for a system to remain in the same state from one observation to the next. For example, if today is rainy, then tomorrow is more likely to be rainy.

However, the use of lagged variables has consistently increased the predictive power of such models. PM 10 is also autocorrelated and persistent, and therefore including lagged PM 10 in the set of explanatory variables strengthens the predictive power of a model [ 24 — 26 ]. Co-pollutants—gases such as nitrogen monoxide NO , nitrogen dioxide NO 2 , carbon monoxide CO and sulphur dioxide SO 2 —have been found to be useful explanatory variables when used in conjunction with meteorological variables [ 24 , 25 , 27 ].

In many countries and especially in urban areas, road transportation is considered to be the largest contributor to PM Road vehicles not only emit exhaust but also resuspend particulate matter [ 28 ].

Where data on traffic are not available, CO and NO x can be used as a proxy for exhaust emissions [ 27 ].

Land usage can also influence PM 10 concentrations, and therefore, land use type may be a useful explanatory variable. One study discovered spatial variations in PM 10 with higher concentrations in commercial areas than in residential and industrial areas [ 29 ]. However, land use classifications are not common in PM 10 models. Another factor affecting PM 10 concentrations is time. Various temporal variables have been used in models of PM 10 concentration.

Variables that reflect the seasonal cycle, such as sine and cosine of Julian day, are important for mean daily PM 10 prediction because they reflect the dry, warm conditions typical in summer and therefore the role of photochemical production in increasing particulate matter concentrations [ 24 ]. Similarly, binary variables are sometimes used to indicate whether a period is cold or warm.

For urban areas variables that reflect diurnal and weekly cycles are important due to high-density commuter and industrial traffic on weekdays contributing significantly to PM 10 levels [ 16 ]. In urban areas of New Zealand PM 10 has distinct diurnal cycles, with peaks between 10 pm and midnight and 8 am and 10 am, which have been found to be independent of population density [ 30 ]. Over the last 10 years, most PM 10 models have included temporal variables.

A very recent approach to estimating PM 10 concentrations is the use of satellite-based remote sensing in addition to ground-level meteorological variables.

One study published in used AOD along with meteorological variables to predict ground-level PM 10 but did not evaluate the degree to which including AOD influenced the outcome of PM 10 predictions [ 32 ].

Regression methods have been used as prediction and estimation tools in a wide range of disciplines including environmental pollution and climate studies. These methods are simple to implement and compute and provide models that are easily interpretable, hence their wide adoption.

Among regression methods, multivariate linear regression MLR is probably the most commonly used statistical method for modelling air pollution and PM Table 1 summarises some recent MLR PM 10 models reported in the literature, highlighting the explanatory variables which contributed to each model. MLR is simply a process of finding a line that best fits a multidimensional cloud of data points. The line of best fit is computed to be the line in which the squared deviations of the observed points from that line are minimised.

This line of best fit provides a model Eq. MLR models are considered to be limited models of PM 10 concentration due to the inability to extend the response to non-central locations of the explanatory variables and to meet the other assumptions of the model [ 14 , 33 ]. Despite possible nonconformity with one or more of the assumptions, MLR has been used extensively for predicting PM 10 and is often used as a benchmark to which other methods are compared.

Much of the PM 10 modelling reported in the literature does not fully provide or explicitly address the data preparation and exploration steps. Thus, it is often difficult to ascertain whether poor performance of MLR models is due to the fact that the data do not meet the assumptions of MLR, and therefore, MLR is a poor choice of model, or that the researchers chose not to refine the model when used only as a benchmark.

In practice the assumption of linearity cannot be confirmed, and linear regression models are considered to be acceptable provided there are only minor deviations from this assumption. PM 10 and its explanatory variables typically do not meet the assumption of linearity [ 22 ].

Often the variables do not have a normal distribution due to the presence of outliers. Issues of linearity, non-normal distribution and homoscedasticity can be addressed by transforming the variable concerned. Such transformations are undertaken to linearise the relationship between the response and explanatory variables making model fitting simpler.

Variable transformations should be handled carefully because in some cases, the transformation can introduce multicollinearity. Hourly PM 10 concentrations in Athens are reported to have a logarithmic distribution so modellers performed a log transform of the PM 10 response variable in order to improve the homoscedasticity of the residuals [ 16 ].

In Chile, maximum hour moving average PM 10 was also found to be logarithmically distributed. Again a log transform was used to normalise the data, and extreme outliers were removed [ 34 ].

After outlier analysis, all data were then normalised to ensure constant variance for each variable. GLM is a generalisation of MLR that relates a linear model to its response variable by a link function and therefore allows for response variables that are not normally distributed. Little difference was observed between the two models, suggesting that the simpler MLR method was a better option than GLM in that case.

Studies in other locales have reported that PM 10 was normally distributed [ 26 ] or that PM 10 concentrations were right skewed [ 14 ]. Studies have found that the use of curvilinear transformations of input variables e. Multicollinearity can be identified by examining the correlations among pairs of explanatory variables.

However, looking at correlations only among pairs of predictors is not the best approach as even when pairwise correlations are small, it is possible that a linear dependence exists among three or more variables.

These metrics are regularly updated to reflect usage leading up to the last few days. Citations are the number of other articles citing this article, calculated by Crossref and updated daily. Find more information about Crossref citation counts. The Altmetric Attention Score is a quantitative measure of the attention that a research article has received online.

Clicking on the donut icon will load a page at altmetric. Find more information on the Altmetric Attention Score and how the score is calculated. Particulate matter is currently measured using particle mass, particle number, and particle size distribution metrics, with other metrics, such as particle surface area, likely to be of increasing importance in the future.

Particulate mass is measured using gravimetric methods, tapered element oscillating microbalances, and beta attenuation instruments and is best suited to use in compliance monitoring, trend analysis, and high spatial resolution measurements.

Particle number concentration is measured by condensation particle counters, optical particle counters, and diffusion chargers. Particle number measurements are best suited to source characterization, trend analysis and ultrafine particle investigations. Particle size distributions are measured by gravimetric impactors, scanning mobility particle sizers, aerodynamic particle sizers, and fast mobility particle sizers.

Particle size distribution measurements are most useful in source characterization and particulate matter property investigations, but most measurement options remain expensive and intrusive.

However, we are on the cusp of a revolution in indoor air quality monitoring and management. Low-cost sensors have potential to facilitate personalized information about indoor air quality IAQ , allowing citizens to reduce exposures to PM indoors and to resolve potential dichotomies between promoting healthy IAQ and energy efficient buildings.

Indeed, the low cost will put this simple technology in the hands of citizens who wish to monitor their own IAQ in the home or workplace, to inform lifestyle decisions. The bar charts show the real-time measured mass concentration bins, measured with an SPS The left chart shows a live measurement of match smoke, clearly richer in smaller particles.

The right chart shows a measurement from Arizona dust, clearly richer in bigger particles. This simple but effective experiment highlights the value of the SPS30 advanced binning feature and the potential for the development of new applications based on particle composition detection.

As mentioned previously a PM sensor is in principle very susceptible to output drift due to the accumulation of dust on the crucial optical parts of the device, namely the laser, the photodiode and the beam-dump used to absorb the laser light and avoid parasitic scattering. Based on more than 20 years of experience in flow sensor design for several demanding markets and applications e. The picture clearly shows that the flow path protects the crucial optical elements from dust exposure, and that the laser and photodiode are completely clean even after the stress test the beam-dump, which is also protected from dust accumulation, is not visible in the photo.

A sensor that works over the whole lifetime of a device guarantees good air quality to the final user and increases energy efficiency and sustainable operation. Your Contact to Sensirion. Contact our sensor experts: Please contact us. Support Center. Please find here an overview of various support topics: Go to Support Center. Find out where we are located: Our Locations.

Share sensitive information only on official, secure websites. JavaScript appears to be disabled on this computer. Please click here to see any active alerts. PM stands for particulate matter also called particle pollution : the term for a mixture of solid particles and liquid droplets found in the air. Some particles, such as dust, dirt, soot, or smoke, are large or dark enough to be seen with the naked eye. Others are so small they can only be detected using an electron microscope.