Scholarlink

<p>Prediction of rainfall is a primary area of interest in the field of meteorology. The assessment of numerical forecasts of rainfall levels and projections of associated climate extremes, such as droughts and floods, is beneficial for enabling timely early warning systems and response measures. The study focuses on West Bengal, chosen for its diverse climatic zones, ranging from the Himalayan foothills to coastal plains, which experience highly variable rainfall patterns due to orographic, local, and cyclonic influences, making it ideal for analyzing rainfall trends and variability. In the present study, rainfall variability, trend and forecasting from 2022 to 2037 have been analysed and 0.25°gridded data were obtained from the Indian Meteorological Department (IMD) using Python. The present study introduces Long Short-Term Memory (LSTM) based on Recurrent Neural Network (RNN) for rainfall prediction. We conducted a time series analysis on monthly rainfall and applied LSTM algorithms on yearly data from 1951 to 2022. Numerous methodologies have been previously suggested for the prediction using statistical analysis, machine learning, and deep learning approaches. Forecasting time series data in meteorology may aid strategic decision-making by organisations tasked with catastrophe avoidance. The neural network is trained and evaluated using the standard rainfall dataset. LSTM models performed well in prediction, with training and testing R<sup>2</sup> values of 0.878 and 0.881, respectively. The root mean square error (RMSE) value for training is 35.15, whereas for testing it is 22.75. The mean absolute error (MAE) value for training is 31.88, and for testing it is 26.62.</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

Analysing and forecasting of rainfall trend and variability of West Bengal using deep learning based LSTM model

  • 期刊论文

  • 发表日期：

    2026年3月25日

查看全文
  • Debasish Mandal,
  • Nawaj Sarif,
  • Sujoy Kumar Malo,
  • Kunal Chakraborty,
  • Snehasish Saha,
  • Padmini Pani

摘要

Prediction of rainfall is a primary area of interest in the field of meteorology. The assessment of numerical forecasts of rainfall levels and projections of associated climate extremes, such as droughts and floods, is beneficial for enabling timely early warning systems and response measures. The study focuses on West Bengal, chosen for its diverse climatic zones, ranging from the Himalayan foothills to coastal plains, which experience highly variable rainfall patterns due to orographic, local, and cyclonic influences, making it ideal for analyzing rainfall trends and variability. In the present study, rainfall variability, trend and forecasting from 2022 to 2037 have been analysed and 0.25°gridded data were obtained from the Indian Meteorological Department (IMD) using Python. The present study introduces Long Short-Term Memory (LSTM) based on Recurrent Neural Network (RNN) for rainfall prediction. We conducted a time series analysis on monthly rainfall and applied LSTM algorithms on yearly data from 1951 to 2022. Numerous methodologies have been previously suggested for the prediction using statistical analysis, machine learning, and deep learning approaches. Forecasting time series data in meteorology may aid strategic decision-making by organisations tasked with catastrophe avoidance. The neural network is trained and evaluated using the standard rainfall dataset. LSTM models performed well in prediction, with training and testing R2 values of 0.878 and 0.881, respectively. The root mean square error (RMSE) value for training is 35.15, whereas for testing it is 22.75. The mean absolute error (MAE) value for training is 31.88, and for testing it is 26.62.