A forecast activity in the lowest layer of the atmosphere, well known for its strongly non-linear physics, is presented in this paper. The forecast method is mainly based on a neural network model, whose structure is briefly described. We stress that preprocessing allows us to extract the main periodicities and to train the network on a residual series of radon data: here the network itself is able to catch the hidden non-linear dynamics. Final results show the ability of the model to predict values of radon concentration and stable layer depth, which represent important physical information for air pollution forecasts near the surface.
Short range forecast of atmospheric radon concentration and stable layer depth by neural network modelling
Pasini A;
2003
Abstract
A forecast activity in the lowest layer of the atmosphere, well known for its strongly non-linear physics, is presented in this paper. The forecast method is mainly based on a neural network model, whose structure is briefly described. We stress that preprocessing allows us to extract the main periodicities and to train the network on a residual series of radon data: here the network itself is able to catch the hidden non-linear dynamics. Final results show the ability of the model to predict values of radon concentration and stable layer depth, which represent important physical information for air pollution forecasts near the surface.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.