World cloud cover feature extraction base on wavelet and statistics from ISCCP D2 dataset

Abstract

In order to extract cloud cover feature from ISCCP D2 dataset, a method of feature extraction using wavelet and statistics was used. This method concerned the characteristic of the cloud cover and the applications requirement, and combined the autocorrelation function, partial autocorrelation function with the wavelet method. We can get the conclusion from the features: (1) the features from wavelet analysis are more evident than the features from original series; (2) most of the cloud amount series in ISCCP D2 dataset are stationary series, and the autocorrelation functions (AF) and partial autocorrelation functions (PAF) shows there are diurnal cycle in these series. As a result, it is possible to establish ARIMA model to estimate the cloud amount for a small region in the world.