In this article, a novel technique based on artificial neural networks (NN) is proposed for cloud coverage short-term forecasting (nowcasting). In particular, the capabilities of multi-layer perceptron NN and time series analysis with nonlinear autoregressive with exogenous input NN
are explored and applied to the European meteorological system ‘Meteosat Second Generation’ with its payload Spinning Enhanced Visible and InfraRed Imager. The general neural architecture consists of a first stage addressing the prediction of the radiance images at six bands (0.6,
0.8, 1.6, 3.9, 6.2 and 10.8 μm). In a second stage a cloud masking algorithm, always based on NN, is applied to the predicted images for the cloud coverage nowcasting. The scheme was compared with the most basic forecast algorithm for the prediction: the persistent model. Two test
areas characterized by different climatology have been considered for the performance analysis. The results show that about 85% of the changes occurring in the time window were recognized by the proposed technique.
No Reference information available - sign in for access.
No Citation information available - sign in for access.
No Supplementary Data.
No Article Media
Document Type: Research Article
Department of Civil Engineering and Computer Science Engineering (D.I.C.I.I.), University of Rome ‘Tor Vergata’, Rome, Italy
GEOMAR Helmholtz Centre for Ocean Research, Kiel, Germany
Remote Sensing & Environmental Modelling Lab, Kiel University, Kiel, Germany
Publication date: December 16, 2016
More about this publication?