Premium
Missing data assessment in a solarimetric network
Author(s) -
Ceballos Juan Carlos,
Braga Célia Campos
Publication year - 1995
Publication title -
international journal of climatology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.58
H-Index - 166
eISSN - 1097-0088
pISSN - 0899-8418
DOI - 10.1002/joc.3370150308
Subject(s) - missing data , standard deviation , principal component analysis , data set , homogeneous , series (stratigraphy) , statistics , time series , eigenvalues and eigenvectors , set (abstract data type) , standard error , data mining , computer science , mathematics , meteorology , geography , geology , paleontology , physics , combinatorics , quantum mechanics , programming language
Principal components analysis was applied to solar irradiation (energy per unit area integrated over a daily period) in Paraiba State, north‐east Brazil, in order to find homogeneous subregions. The network data set included measurements of 19 solarimetric stations, generated by bimetallic Robitzsch‐type actinographs with associated daily errors not greater than 5 per cent. Two months were considered (February and August 1976). A simple procedure is developed for assessing missing data in time series of any station in the network, based on the remaining known data and on a small number of network eigenvectors. If applied over a homogeneous subregion, a specific new principal components analysis leads to a lower dimensioned system with even fewer eigenvectors used for proper assessment of missing data. Standard deviation of errors is lower than half the data standard deviation (the same order of expected instrumental errors). The method may be adapted easily in order to simulate (synthetize) time series of a site with no station, provided that time correlation fields vary smoothly within the region.