Open AccessLong‐range forecast of Indian summer monsoon rainfall using an artificial neural network model
Author(s) -
Acharya Rajashree,
Pal Jayanti,
Das Debanjana,
Chaudhuri Sutapa
Publication year - 2019
Publication title -
meteorological applications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.672
H-Index - 59
eISSN - 1469-8080
pISSN - 1350-4827
DOI - 10.1002/met.1766
Subject(s) - artificial neural network , multilayer perceptron , anomaly (physics) , range (aeronautics) , meteorology , data set , robustness (evolution) , climatology , computer science , environmental science , artificial intelligence , geology , geography , engineering , biochemistry , chemistry , physics , aerospace engineering , gene , condensed matter physics
This study develops an artificial neural network (ANN) model with a nonlinear perceptron rule for use in the long‐range forecasting (LRF) of Indian summer monsoon rainfall (ISMR). In developing the model, two predictor sets are adopted from the India Meteorological Department (IMD), SET‐I and SET‐II, to prepare the input matrix of the model, while the output is ISMR. The data used were collected over the period 1980–2017. The model is trained with input data from 1980 to 2012, and the skill of the model is estimated by validating the model output with observation during the period 2013–2017. The result reveals that that second‐stage forecast is better than first‐stage forecast due to the incorporation of a North Atlantic sea surface pressure anomaly and a North Central Pacific zonal wind anomaly at 850 hPa in the input matrix. The study further reveals that the multilayer perceptron (MLP) model with a back‐propagation algorithm is best among the ANN models used in the study. The prediction capability of the ANN model is also checked by comparing it with a multiple nonlinear regression (MNLR) model developed with the two predictor sets. The robustness of the prediction accuracy is estimated by computing Willmott's index for each of the ANN and MNLR models.