Scholarlink

<p>In locations with insufficient surface water, groundwater is even more crucial for domestic, agricultural, and industrial applications. India, the world's largest groundwater consumer, struggles with groundwater management due to climate change and human activity. This work uses Raipur, Chhattisgarh, India, as a case study to develop and test deep learning models for groundwater level forecasting. Recurrent Neural Networks (RNNs), Bidirectional Long Short-Term Memory (Bi-LSTM), and Long Short-Term Memory (LSTM) were utilized to predict groundwater levels year-round. The study uses a dataset spanning 51&#xa0;years, from March 1973 to January 2024. After training on 80% of the data and testing on 20%, hyperparameters were fine-tuned to optimize performance. Evaluation measures included R-squared, cosine similarity, root-mean-squared error, mean absolute error, and mean absolute percentage error. The results showed that the Bi-LSTM model predicted groundwater levels better than the LSTM and RNN models. Bi-LSTM improved forecast accuracy by incorporating context from past and future conditions due to its bidirectional processing capacity. The study stresses the impact of droughts, monsoon rains, and agricultural and industrial groundwater extraction on groundwater levels. Integrating these components helps illuminate groundwater fluctuations in models. Accurate projections enable sustainable water resource management by improving planning and decision-making. This study provides a solid foundation for using advanced deep learning approaches to anticipate groundwater levels in other regions with similar challenges.</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

Predicting Groundwater Levels Using Advanced Deep Learning Models: A Case Study of Raipur, India

  • 期刊论文

  • 发表日期：

    2025年5月13日

查看全文
  • Sneha Thakur,
  • Sanjeev Karmakar

摘要

In locations with insufficient surface water, groundwater is even more crucial for domestic, agricultural, and industrial applications. India, the world's largest groundwater consumer, struggles with groundwater management due to climate change and human activity. This work uses Raipur, Chhattisgarh, India, as a case study to develop and test deep learning models for groundwater level forecasting. Recurrent Neural Networks (RNNs), Bidirectional Long Short-Term Memory (Bi-LSTM), and Long Short-Term Memory (LSTM) were utilized to predict groundwater levels year-round. The study uses a dataset spanning 51 years, from March 1973 to January 2024. After training on 80% of the data and testing on 20%, hyperparameters were fine-tuned to optimize performance. Evaluation measures included R-squared, cosine similarity, root-mean-squared error, mean absolute error, and mean absolute percentage error. The results showed that the Bi-LSTM model predicted groundwater levels better than the LSTM and RNN models. Bi-LSTM improved forecast accuracy by incorporating context from past and future conditions due to its bidirectional processing capacity. The study stresses the impact of droughts, monsoon rains, and agricultural and industrial groundwater extraction on groundwater levels. Integrating these components helps illuminate groundwater fluctuations in models. Accurate projections enable sustainable water resource management by improving planning and decision-making. This study provides a solid foundation for using advanced deep learning approaches to anticipate groundwater levels in other regions with similar challenges.