Premium
Forecasting all‐India summer monsoon rainfall using regional circulation principal components: a comparison between neural network and multiple regression models
Author(s) -
Can Alex J.,
McKendry Ian G.
Publication year - 1999
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/(sici)1097-0088(19991130)19:14<1561::aid-joc434>3.0.co;2-3
Subject(s) - climatology , monsoon , geopotential height , linear regression , principal component analysis , artificial neural network , environmental science , regression , resampling , meteorology , precipitation , geography , mathematics , statistics , computer science , geology , machine learning
Abstract Pre‐monsoon principal components (PCs) of circulation fields covering the South Asian subcontinent were used as predictors for all‐India summer monsoon rainfall (AISMR) over the period 1958–1993. Predictive skill of non‐linear neural network models and linear multiple regression models was compared using a bootstrap‐based resampling procedure. Monsoon precursor signals represented by PCs were investigated and comparisons made with a recent observational and general circulation modelling study. Pre‐monsoon PCs of the 200 hPa geopotential height field in May formed a compact, interpretable, and significant set of predictors for AISMR. Predictive skill was comparable to or better than that reported in prior modelling studies, each of which used optimized sets of regional and global predictors. No improvement was noted when using data from multiple atmospheric levels, and skill at lead times more than 1 month prior to monsoon onset in June was poor. For May predictors there were only small differences in skill between the neural network and multiple regression models, although the neural network results at longer lead times tended to be better than those shown by multiple regression. Interestingly, the 850 hPa PCs in January showed a maximum in predictive skill that was only evident in the neural network model results. The strength of this relationship suggests that further investigation into the use of January 850 hPa predictors for the long‐range forecasting of AISMR is warranted. Copyright © 1999 Royal Meteorological Society