Can we predict atmospheric PM2.5 concentration more accurately?
Air pollution is a major concern in many counties. Air pollution levels are usually determined by the concentrations of air pollutants such as nitrogen dioxide, sulphur dioxide, carbon monoxide, ozone and particulate matters (PMs). Meanwhile, air pollution factors, such as traffic flow and unfavoura...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Conference Paper |
| Published: |
SETAC AP
2012
|
| Online Access: | http://hdl.handle.net/20.500.11937/23174 |
| _version_ | 1848751076813045760 |
|---|---|
| author | Jian, Le Chan, Kit Yan |
| author2 | Koji Arizono |
| author_facet | Koji Arizono Jian, Le Chan, Kit Yan |
| author_sort | Jian, Le |
| building | Curtin Institutional Repository |
| collection | Online Access |
| description | Air pollution is a major concern in many counties. Air pollution levels are usually determined by the concentrations of air pollutants such as nitrogen dioxide, sulphur dioxide, carbon monoxide, ozone and particulate matters (PMs). Meanwhile, air pollution factors, such as traffic flow and unfavourable meteorological factors, can either be the pollution sources or may affect the formation and growth of air pollutants and the ability of the atmosphere to disperse air pollutants. As air pollution can result in acute, chronic diseases, or even be life threatening, it is essential to develop propitiate models to predict air pollution levels and determine the contributions from air pollution factors. In recent decade, artificial neural network (ANN) is rapidly emerged into environmental science as an advanced technology in forecasting air pollution levels. However, ANN has a “black-box” feature and is unable to provide explicit knowledge of significant air pollution factors that contribute to air pollution levels. In order to overcome this limitation, we developed a neural network based knowledge discovery system and reported below. The new system consists of two units: (1) an ANN unit to estimate air pollutant (e.g.PM2.5) concentrations based on relevant pollution factors (traffic flow, wind speed, temperature, relative humidity, and barometric pressure), and (2) a knowledge discovery unit to extract explicit knowledge from the ANN unit. Survey data on mass concentrations of PM2.5, meteorological and traffic data measured near a busy traffic road in Hangzhou, China were applied to demonstrate the practicability of the system.Fifteen cross validations were conducted and the results showed that the new neural network based knowledge discovery system can yield smaller validation errors than the regression model. Based on the main effect analysis via the new knowledge discovery unit, traffic flow provides more significant contribution to the predicted concentrations of PM2.5 than these meteorological factors. In conclusion, this new ANN based knowledge discovery system can predict air pollution level (PM2.5) more accurately and identify significant contributors to air pollution levels. The system has potential application in planning air monitoring programs to achieve cost-effective outcomes. |
| first_indexed | 2025-11-14T07:46:59Z |
| format | Conference Paper |
| id | curtin-20.500.11937-23174 |
| institution | Curtin University Malaysia |
| institution_category | Local University |
| last_indexed | 2025-11-14T07:46:59Z |
| publishDate | 2012 |
| publisher | SETAC AP |
| recordtype | eprints |
| repository_type | Digital Repository |
| spelling | curtin-20.500.11937-231742017-02-28T01:37:24Z Can we predict atmospheric PM2.5 concentration more accurately? Jian, Le Chan, Kit Yan Koji Arizono Air pollution is a major concern in many counties. Air pollution levels are usually determined by the concentrations of air pollutants such as nitrogen dioxide, sulphur dioxide, carbon monoxide, ozone and particulate matters (PMs). Meanwhile, air pollution factors, such as traffic flow and unfavourable meteorological factors, can either be the pollution sources or may affect the formation and growth of air pollutants and the ability of the atmosphere to disperse air pollutants. As air pollution can result in acute, chronic diseases, or even be life threatening, it is essential to develop propitiate models to predict air pollution levels and determine the contributions from air pollution factors. In recent decade, artificial neural network (ANN) is rapidly emerged into environmental science as an advanced technology in forecasting air pollution levels. However, ANN has a “black-box” feature and is unable to provide explicit knowledge of significant air pollution factors that contribute to air pollution levels. In order to overcome this limitation, we developed a neural network based knowledge discovery system and reported below. The new system consists of two units: (1) an ANN unit to estimate air pollutant (e.g.PM2.5) concentrations based on relevant pollution factors (traffic flow, wind speed, temperature, relative humidity, and barometric pressure), and (2) a knowledge discovery unit to extract explicit knowledge from the ANN unit. Survey data on mass concentrations of PM2.5, meteorological and traffic data measured near a busy traffic road in Hangzhou, China were applied to demonstrate the practicability of the system.Fifteen cross validations were conducted and the results showed that the new neural network based knowledge discovery system can yield smaller validation errors than the regression model. Based on the main effect analysis via the new knowledge discovery unit, traffic flow provides more significant contribution to the predicted concentrations of PM2.5 than these meteorological factors. In conclusion, this new ANN based knowledge discovery system can predict air pollution level (PM2.5) more accurately and identify significant contributors to air pollution levels. The system has potential application in planning air monitoring programs to achieve cost-effective outcomes. 2012 Conference Paper http://hdl.handle.net/20.500.11937/23174 SETAC AP restricted |
| spellingShingle | Jian, Le Chan, Kit Yan Can we predict atmospheric PM2.5 concentration more accurately? |
| title | Can we predict atmospheric PM2.5 concentration more accurately? |
| title_full | Can we predict atmospheric PM2.5 concentration more accurately? |
| title_fullStr | Can we predict atmospheric PM2.5 concentration more accurately? |
| title_full_unstemmed | Can we predict atmospheric PM2.5 concentration more accurately? |
| title_short | Can we predict atmospheric PM2.5 concentration more accurately? |
| title_sort | can we predict atmospheric pm2.5 concentration more accurately? |
| url | http://hdl.handle.net/20.500.11937/23174 |