Scholarlink

<p>This study presents a comparative analysis of deep learning models for predicting groundwater levels (GWL) in coastal aquifers, focusing on sedimentary and metamorphic aquifers in Nova Scotia. Using extensive datasets from the sedimentary aquifer in Sydney (1984–2022) and the metamorphic aquifer in Yarmouth (1990–2022), the research evaluates the performance of models including Recurrent Neural Networks (RNNs), Long Short-Term Memory networks (LSTMs), Convolutional Neural Networks (CNNs), Gated Recurrent Units (GRUs), and Multilayer Perceptrons (MLPs). Six different combinations of hydrological and environmental inputs were tested, with combination 5, comprising GWL from the past 15&#xa0;months, nearby observation groundwater well data (Next-GWL), sea level (SL), air temperature (T), rain (R), and snow (S), showing the highest predictive accuracy. The models were evaluated over three forecast horizons: 6&#xa0;months (short term), 12&#xa0;months (medium term), and 24&#xa0;months (long term). Performance metrics such as the Coefficient of Correlation (R), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Nash–Sutcliffe Efficiency (NSE) were used to assess model accuracy. Notably, LSTM and GRU models were highly adaptable, capturing temporal patterns effectively. For example, higher correlation coefficients and Nash–Sutcliffe efficiency are generally observed in the training phases, with the LSTM model combination 5 reaching an R value of 0.95 for metamorphic aquifers and 0.92 for sedimentary ones. The study underscores the importance of integrating multiple environmental variables and using advanced deep learning techniques to improve groundwater level predictions, offering valuable insights for water resource management in coastal regions affected by climate change and sea-level rise.</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 in coastal aquifers using deep learning models: a comparative study of sedimentary and metamorphic aquifers in nova scotia

  • 期刊论文

  • 发表日期：

    2025年5月3日

查看全文
  • Saeideh Samani,
  • Meysam Vadiati,
  • Ozgur Kisi

摘要

This study presents a comparative analysis of deep learning models for predicting groundwater levels (GWL) in coastal aquifers, focusing on sedimentary and metamorphic aquifers in Nova Scotia. Using extensive datasets from the sedimentary aquifer in Sydney (1984–2022) and the metamorphic aquifer in Yarmouth (1990–2022), the research evaluates the performance of models including Recurrent Neural Networks (RNNs), Long Short-Term Memory networks (LSTMs), Convolutional Neural Networks (CNNs), Gated Recurrent Units (GRUs), and Multilayer Perceptrons (MLPs). Six different combinations of hydrological and environmental inputs were tested, with combination 5, comprising GWL from the past 15 months, nearby observation groundwater well data (Next-GWL), sea level (SL), air temperature (T), rain (R), and snow (S), showing the highest predictive accuracy. The models were evaluated over three forecast horizons: 6 months (short term), 12 months (medium term), and 24 months (long term). Performance metrics such as the Coefficient of Correlation (R), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Nash–Sutcliffe Efficiency (NSE) were used to assess model accuracy. Notably, LSTM and GRU models were highly adaptable, capturing temporal patterns effectively. For example, higher correlation coefficients and Nash–Sutcliffe efficiency are generally observed in the training phases, with the LSTM model combination 5 reaching an R value of 0.95 for metamorphic aquifers and 0.92 for sedimentary ones. The study underscores the importance of integrating multiple environmental variables and using advanced deep learning techniques to improve groundwater level predictions, offering valuable insights for water resource management in coastal regions affected by climate change and sea-level rise.