A deep and wide neural network to predict summer monsoon rainfall using time series data. (18th January 2023)

• Record Type: Journal Article
• Authors: Bajpai, Vikas; Bansal, Anukriti; Dash, Subrat
• Is Part Of: Concurrency and computation. Volume 35:Number 8(2023)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2023-01-18
• Subjects: convolutional neural network -- long and short term memory -- multilayer perceptron -- summer monsoon rainfall prediction -- time‐series forecasting -- wide and deep neural network; Parallel processing (Electronic computers) -- Periodicals; Parallel computers -- Periodicals; 004.35
• DOI: 10.1002/cpe.7626
• Languages: English
• ISSNs: 1532-0626

Abstract

Summary: The importance of monsoon rainfall cannot be ignored as it affects round‐the‐year activities ranging from agricultural to industrial. Due to the nonlinear and dynamic nature of monsoon rainfall, its accurate forecast becomes a challenging task. In this paper, we propose a novel deep and wide neural network model to predict summer monsoon rainfall using historical time‐series data of 118 years. We have considered Rajasthan as our area of study, which is the largest state of India with varying atmospheric regions. Two different types of datasets are used: gridded rainfall dataset from the Indian Meteorological Department (IMD) and station rainfall data from Water Resources Department (WRD). Results obtained on both the datasets demonstrate that a single model works well in efficiently predicting summer monsoon rainfall in different geographical regions of Rajasthan, ranging from plains and plateaus to deserts and hills. We compare the results with various advanced deep learning algorithms such as 1‐dimensional convolutional neural network, long short term memory, and multilayer perceptron. The comparison results show that the proposed method clearly outperforms these algorithms in terms of its forecasting ability, which is evaluated using root mean square error and mean absolute error.