Journal
INTERNATIONAL JOURNAL OF FORECASTING
Volume 37, Issue 1, Pages 58-71Publisher
ELSEVIER
DOI: 10.1016/j.ijforecast.2020.03.001
Keywords
Stacked autoencoder; Automated feature learning; Predictor identification; Monsoon prediction; Ensemble regression model; Indian summer monsoon
Categories
Ask authors/readers for more resources
This study explores the use of deep learning for feature reduction and discovery of predictors for Indian summer monsoon rainfall. The combination of climatic variables as predictors outperforms individual variables, with the model being able to accurately predict the monsoon two months in advance.
The study of climatic variables that govern the Indian summer monsoon has been widely explored. In this work, we use a non-linear deep learning-based feature reduction scheme for the discovery of skilful predictors for monsoon rainfall with climatic variables from various regions of the globe. We use a stacked autoencoder network along with two advanced machine learning techniques to forecast the Indian summer monsoon. We show that the predictors such as the sea surface temperature and zonal wind can predict the Indian summer monsoon one month ahead, whereas the sea level pressure can predict ten months before the season. Further, we also show that the predictors derived from a combination of climatic variables can outperform the predictors derived from an individual variable. The stacked autoencoder model with combined predictors of sea surface temperature and sea level pressure can predict the monsoon (June-September) two months ahead with a 2.8% error. The accuracy of the identified predictors is found to be superior to the state-of-the-art predictions of the Indian monsoon. (C) 2020 International Institute of Forecasters. Published by Elsevier B.V. All rights reserved.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available