Scholarlink

<p>The Indian economy is heavily influenced by its agricultural sector, which critically depends on rainfall patterns. The average seasonal rainfall, based on historical data from 1961 to 2018, is around 88&#xa0;cm with a variation coefficient of about 10%. Years with rainfall exceeding 110% of this average are considered excess, while those with less than 90% are categorized as drought years. Traditional methods for forecasting monsoon precipitation have often been inadequate in providing precise quantitative analyses for the entire country. This has spurred ongoing research aimed at developing more reliable forecasting techniques. In this paper, we introduce a novel forecasting method to predict the southwest Indian summer monsoon rainfall (ISMR) by leveraging an artificial neural networks approach that utilizes ten climate parameters. This method involves pre-processing the ISMR data to remove anomalies. Additionally, an artificial neural network-based autoencoder is employed to recreate the input vector using a stacked auto-encoder, which captures intricate data features and reduces data dimensionality. To recreate the sea surface pressure (SLP) input parameter, the auto-encoder identifies the six major correlated SLP vectors for machine learning model training and forecasting using a 16-year sliding window. The vectors of actual observed ISMR and model-forecasted rainfall were found to correlate with a Pearson correlation coefficient of 0.829. An analysis of the percentage deviation of ISMR data from the seasonal mean rainfall for the period 1991–2019 highlights recent deficit, normal, and excess monsoon rainfall years. Monthly statistical analyses of historical Indian rainfall data are conducted across 36 meteorological subdivisions, seven homogeneous regions, and the country as a whole.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Forecasting of Southwest Indian Summer Monsoon Rainfall Using Artificial Neural Networks

  • 期刊论文

  • 发表日期：

    2025年11月13日

查看全文
  • Aman Kumar,
  • Samayveer Singh,
  • Aruna Malik,
  • Manju

摘要

The Indian economy is heavily influenced by its agricultural sector, which critically depends on rainfall patterns. The average seasonal rainfall, based on historical data from 1961 to 2018, is around 88 cm with a variation coefficient of about 10%. Years with rainfall exceeding 110% of this average are considered excess, while those with less than 90% are categorized as drought years. Traditional methods for forecasting monsoon precipitation have often been inadequate in providing precise quantitative analyses for the entire country. This has spurred ongoing research aimed at developing more reliable forecasting techniques. In this paper, we introduce a novel forecasting method to predict the southwest Indian summer monsoon rainfall (ISMR) by leveraging an artificial neural networks approach that utilizes ten climate parameters. This method involves pre-processing the ISMR data to remove anomalies. Additionally, an artificial neural network-based autoencoder is employed to recreate the input vector using a stacked auto-encoder, which captures intricate data features and reduces data dimensionality. To recreate the sea surface pressure (SLP) input parameter, the auto-encoder identifies the six major correlated SLP vectors for machine learning model training and forecasting using a 16-year sliding window. The vectors of actual observed ISMR and model-forecasted rainfall were found to correlate with a Pearson correlation coefficient of 0.829. An analysis of the percentage deviation of ISMR data from the seasonal mean rainfall for the period 1991–2019 highlights recent deficit, normal, and excess monsoon rainfall years. Monthly statistical analyses of historical Indian rainfall data are conducted across 36 meteorological subdivisions, seven homogeneous regions, and the country as a whole.