| Links |
 |
 |
|
|||||||||||||||
|
URBAN
AIR POLLUTION: DATA INTERPRETATION AND NOWCASTING BY NEURAL NETWORKS The problem of atmospheric air pollution is especially worrying in urban areas where the output of pollutants is high and the number of people exposed to health hazards grows constantly. Some years ago, Italian legislation made it compulsory for Italian cities to equip themselves with an air monitoring network. In the Mestre (Venice, Italy) area, where some air pollution monitoring instruments had already been in operation for some time, mainly to monitor industrial pollution, the need to abide by the new legislation led to reconstructing of the already available monitoring network and to widen their scope so as to include the monitoring of road traffic pollution. This work is based on the data provided from the Venice Municipality, as well as on some meteorological parameters and vehicular traffic measures. Its aimed to assess air quality and to implement models for the monitoring of severe pollution episodes. The greatest attention is devoted to carbon monoxide, to sulphur acids and to ozone. CO and NO have been considered as primary pollutants and, consequently, as reliable indicators of the presence and type of emission sources. On the other hand, O3 and NO2 were zeroed in on to study photochemical smog, a secondary type of very noxious pollution upon which many international studies are currently focusing. Forecasting models have been implemented using Neural Networks, a product of artificial intelligence that is currently been applied to a growing number of scientific researches. Neural Networks won over other more traditional approaches (like deterministic and traditional stochastic models) due to their being more suitable at a structural level to process available data (very long time series and multiparameter but punctual data). In the economy of the work, data retrieval was of the same importance of their successive elaboration and interpretation. The information requested for a serious environmental research are indeed frequently scattered and rarely co-ordinated in well organised and exhaustive data bases. Chapter 1. Notes about the Italian legislation on urban air pollution Air Quality Standards Monitoring Networks Attention and alarm levels Some crux point Relationship between the Italian and the European Community legislation. Chapter
2. Primary
and secondary pollutants Photochemical smog Air pollution meteorology
Meteorological parameters and their measurement Urban environment.
Heat island Polluting sources Main pollutant toxicity Intervention
outlooks. Chapter
3. Models
for the study of air quality: Neural Networks Structure and running
of neural networks The activation phase The learning phase The generalisation
phase Vantages and limitations of Neural Networks. Application of
Neural Networks to the Statistical Analysis. Chapter
4.
The landscape Traffic Characterisation of the Metre area micro-climate
Air pollution monitoring The monitoring network of Ente Zona di
Porto Marghera The State monitoring network. Chapter
5. Statistical
data analysis The data Descriptive statistics Hourly and daily stratified
data analysis The correlation matrix Time series Fourier analysis
Acute episodes analysis. Chapter
6.
Nowcasting of Mestre air quality with Neural Network Ozone Models
Carbon monoxide models sulphur dioxide models Conclusions.
|