Air pollution occurs when the surrounding air contains gases, dust, fumes or odor in high enough quantities to be harmful. That is, amounts that can be harmful to the health of humans and animals or enough to cause damage to plants and materials.
Pollution is often measured as an air quality index (AQI) rating and is used by a number of government agencies to communicate the extent of air pollution in an area. As the AQI increases, the percentage of the population likely to experience adverse health effects also increases.
People’s awareness of air pollution is rapidly increasing and as such, so is the demand for accurate air quality forecasts (Saadi et al. 2005).
Forecasting pollution however, like weather, is difficult to predict. There are a huge number of variables to take into account - some of which are rather “unpredictable,” such as government intervention and natural disasters, both of which can have a tremendous impact on AQI (Dye, 2003).
Nonetheless, accurate air quality forecasts are becoming increasingly important tools that can provide significant societal and economic benefits - the greatest of which is planning (Saadi et al. 2005). Forecasts enable people to take precautionary measures to avoid or limit exposure to unhealthy levels of airborne pollutants (Dye, 2003). Governments can also make use of early forecasting to establish procedures early on, that could help to reduce the severity of local pollution levels (Saadi et al.2005).
Factors that Influence Pollution Levels
Pollution shows strong correlations with local weather conditions and nearby emissions however, it’s becoming increasing well documented that long-range transport of pollution is another significant influencing factor in local AQI readings (NARSTO, 2003). Predicting air quality, therefore, not only involves the difficulties of weather forecasting, it also requires knowledge of pollutant concentrations and emissions from both surrounding and distant locations, taking into account movement and possible transformations (Saadi et al. 2005).
When looking at the array of factors that can influence AQI forecasting you come to the understanding that forecasting can be both subjective and objective. To ensure the most accurate forecasting predictions we need to grow the database - the larger the database the greater the potential for accurate predictions are. To improve accuracy forecasting programs also make use of continuous real-time data, which is then used to develop forecasting methods, monitor current conditions, and evaluate forecasting performances and amend forecasts accordingly (Ballagas et al. 2003)
Air Quality-forecasting techniques
Meteorological forecasting, or weather forecasting, is cornerstone to all of the techniques described below. Step one to an accurate air quality forecast is obtaining or creating an excellent weather forecast. Weather forecasts make use of both science and technology to make predictions. As a general rule, the greater the combination of methods used in a prediction the more accurate the forecast will be
Forecasting is a subjective and objective technique that contains a large variety of techniques, ranging from the simple to the complex. Forecasts generally range from a period of one to three days – the further the forecast generally the less accurate it will be. For simplicity sake, methods are categorized into three main categories: (U.S. EPA, 1999):
- Statistical Methods:
- Classification and Regression Tree (CART) – This statistical method was designed to classify data into dissimilar groups. Using specialized software it identifies variables (meteorological or air quality) that are strongly correlated with ambient pollution levels. This data is then used to create a decision tree that forecasts concentrations based on predictive variables (weather conditions) and their level of correlation with pollutant concentrations.
- Regression analysis – Regression analysis estimates relationships between variables. By analyzing historical data sets, we are able to determine associations between pollution levels and meteorological data variables. This results in a multivariate linear regression equation that can be used to forecast future pollution levels.
- Artificial Neural Networks - Historical data can be analyzed in a more complex way. Artificial Neural Networks calculate the association between historical data and atmospheric factors through the application of adaptive learning and pattern recognition techniques. This method uses computer-based algorithms that are designed to simulate the human brain in terms of pattern recognition. Making use of complicated non-linear data, identify “unforeseeable” trends in data. This is debatably the most suited method for forecasting pollution due to its multi-dimensional approach
- Three Dimensional (3-D) Models
- Emission models – These models simulate the spatial distribution of emissions and time-dependent emissions of pollutants from both natural and anthropogenic sources.
- Meteorological models – These models forecast all meteorological conditions that determines the transport, deposition, mixing, chemistry, and the emission of pollutants. The model then predicts the ambient levels of pollution by creating a trajectory model - making use of the 3-D meteorological model and the emissions data previously collected.
- Chemical models – chemical models look at the transformation of primary (emitted) pollution into secondary pollution, including their composition and morphology. Using fundamental chemical laws, spectroscopic properties (UV rays), and thermodynamic relationships the model determines a pollutants final composition and morphology.
- Lagrangian models make use of meteorological field data to show the transport and dispersion of pollutants as individual air pockets over time. This results in a computationally efficient network. One problem with this method, however, is that air pollution often involves non-linear chemistry making it hard to characterize the interaction of a large number of individual sources.
- Eulerian models make use of fixed grids, both vertically and horizontally. All chemical equations are solved simultaneously in the grids including the exchange of pollutants between cells. Different grids are used depending on the conditions and requirements. Course grids are generally used over rural areas (homogenous regions) and finer grids will be used over urban areas (heterogeneous). These models are able to produce three-dimensional concentration fields for several pollutants but require significant computational power and expertise.
Climatology is based on the assumption that the past is a relatively good indicator for the future. This method is based on the relationship between specific meteorological conditions and pollution levels, and therefore can be very one-dimensional. This method is often extended to include weather patterns - matching weather patterns to pollution patterns.
Climatology as a forecasting method is often seen as a tool to complement other forecasting methods. This is largely due to the limitations of the method - which include difficulties in predicting abrupt changes in emission patterns, and requiring a large quantity of data to be able to establish realistic trends.
The association between air quality and specific meteorological parameters can be quantified using a variety of statistical techniques. The three of most commonly used methods are as follows:
One disadvantage of the above statistical methods is that they assume stability in terms of the processes that affect air quality. Therefore, any drastic changes to emissions or climate (short or long term) will severely diminish the accuracy of these techniques. There are more complex methods, however, that attempt to take these shortfalls into account – the three-dimensional models.
This is a deterministic approach to the prediction of air quality; it creates 3-D Models that seek to mathematically represent all the important processes that have an impact on ambient pollution levels. The model simulates the emission, transport and transformation of air pollution by making use of several sub models within the hub of the model. Examples of sub models are:
Classifications of three-dimensional models are based on the methods used to simulate the distribution of pollution concentrations. They are classified as either Lagrangian or Eulerian:
To ensure that forecasting is as accurate as possible we need to ensure that the air quality forecasting system contains an arrangement of compatible components. These will include predictor values/techniques and observed networks that are capable of providing real time measurements of the atmospheric condition. These measures are used to create the models and evaluate the quality of the forecast.
Air Quality forecasts are predictions and by nature, inexact – as are weather forecasts. Although they may frequently be accurate - at times variations will occur due to the unpredictable nature of air pollution. Two particularly unpredictable events include natural disasters and abrupt changes in emissions (factories reducing emission due to certain events, such as government sanctioned clean air).
Forecasting techniques are rapidly improving and will continue to become more accurate in the future. Governments and the public understand the need for air quality forecasts not only to monitor their own health and safety but also for governments to apply adaptive management techniques to try and reduce air pollution. Raising public awareness about air pollution is of key importance and AQI forecasts are a way of raising awareness through providing information.
Al-Saadi, J., Szykman, J., Pierce, R. B., Kittaka, C., Neil, D., Chu, D. A., ... & Fishman, J. (2005). Improving national air quality forecasts with satellite aerosol observations. Bulletin of the American Meteorological Society, 86(9), 1249-1261.
Ballagas, M. R., Bishop, M. M., Bridgers, M. G., Browner, M. R., Carlson, M. C., Casmassi, M. J., ... & Weiss, M. Guidelines for Developing an Air Quality (Ozone and PM2 5) Forecasting Program.
Dye, T. S. (2003). Guidelines for developing an air quality (Ozone and PM2. 5) forecasting program.
Laumbach, R., Meng, Q., & Kipen, H. (2015). What can individuals do to reduce personal health risks from air pollution?. Journal of thoracic disease, 7(1), 96.
NARSTO, cited (2003): Particulate matter science for policy makers: A NARSTO assessment. [Available online at www.cgenv.com/Narsto/.]
U. S. Environmental Protection Agency (EPA), (1999). ftp:/www.epa.gov/pub/scram001/modelingcenter/NOxSIPcall/emissions
National Science and Technology Council (U.S.). Air Quality Research Subcommittee. (2001). Air quality forecasting: a review of federal programs and research needs. Boulder, Colo.: NOAA Aeronomy Laboratory