Premium
A comparison of three predictor selection methods for statistical downscaling
Author(s) -
Yang Chunli,
Wang Ninglian,
Wang Shijin
Publication year - 2016
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.4772
Subject(s) - downscaling , precipitation , environmental science , climatology , selection (genetic algorithm) , statistical analysis , statistics , regression analysis , stepwise regression , mathematics , meteorology , computer science , geography , geology , artificial intelligence
Three predictor selection methods [correlation analysis, partial correlation analysis and stepwise regression analysis (SRA)] that are commonly used for statistical downscaling are compared in terms of the uncertainty assessments of their downscaled results using the same statistical downscaling model (SDSM). Uncertainty is assessed by comparing several statistical indices for observed and downscaled daily precipitation, daily maximum and minimum temperature, monthly means and variances of daily precipitation and daily temperature. Besides these, the distributions of monthly mean of daily precipitation, monthly dry and wet days also are considered. The analysis employs the SDSM and 54 years (1961–2014) of observed daily precipitation and temperature together with National Center for Environmental Prediction ( NCEP ) reanalysis predictors. A comparison of the different methods for selecting predictors indicates that SRA is slight better than other two methods in most statistical indices.