Short-range visibility forecast by means of neural-network modelling: a case-study |
| |
Authors: | A Pasini and S Potestà |
| |
Institution: | (1) Servizio Meterologico dell’Aeronautica, 2o CMR Aeroporto ?De Bernardi?, Via di Pratica di Mare, I-00040 Pratica di Mare (Roma), Italy |
| |
Abstract: | Summary A neural-network approach to a hetero-associative prognostic problem in the field of meterology is presented. In particular,
the application to the short-range visibility forecast of a neural back-propagation model with adaptive training and test
procedures is considered. Our case-study leads to the following results: we achieve a better theoretical understanding of
the behaviour of neural networks handling meteorological data (for example, the phenomenon of ?network overfitting? and a
seasonal-change influence on the performance of the model are recognised); the forecast performances of the network are always
better than the persistence and subjective forecasts (in particular, the forecast of the visibility tendency is very good). |
| |
Keywords: | Meteorology |
本文献已被 SpringerLink 等数据库收录! |
|