Scholarlink

<p>Groundwater is vital for sustaining life, particularly in regions facing water scarcity. Effective management of groundwater resources requires accurate mapping of potential groundwater zones. This research incorporates Convolutional Neural Networks (CNN) to map precisely Groundwater Potential (GWP) zones in Mangaluru, a coastal taluk in Karnataka, India. Suitability of ten GWP conditioning factors: Elevation, Slope, Aspect, Rainfall, Geology, Geomorphology, Soil, Land Use and Land Cover (LULC), Drainage Density, and Topographic Wetness Index (TWI) is considered using Multicollinearity analysis. The CNN model performance was compared with Multilayer Perceptron (MLP) and Long Short-Term Memory (LSTM) models, and it outperformed by achieving an overall accuracy of 94.23% and an Area under the Receiver Operating Characteristic (AUC-ROC) of 94%. The resulting GWP map was classified into three zones: high (74.98%), moderate (17.13%), and low (7.88%) potential. Validation using groundwater level data from twenty-nine monitoring wells yielded an accuracy of 77%. The findings demonstrate the effectiveness of CNN for GWP mapping and provide valuable insights for sustainable groundwater resource management, policy and decision-making.</p> Graphical Abstract <p></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

Groundwater Potential Mapping for Mangaluru in India, a Coastal Urban Environment using Convolutional Neural Networks

  • 期刊论文

  • 发表日期：

    2025年12月3日

查看全文
  • Subrahmanya Kundapura,
  • Apoorva Kowdley Venkatesh,
  • Umashankar Kandpal

摘要

Groundwater is vital for sustaining life, particularly in regions facing water scarcity. Effective management of groundwater resources requires accurate mapping of potential groundwater zones. This research incorporates Convolutional Neural Networks (CNN) to map precisely Groundwater Potential (GWP) zones in Mangaluru, a coastal taluk in Karnataka, India. Suitability of ten GWP conditioning factors: Elevation, Slope, Aspect, Rainfall, Geology, Geomorphology, Soil, Land Use and Land Cover (LULC), Drainage Density, and Topographic Wetness Index (TWI) is considered using Multicollinearity analysis. The CNN model performance was compared with Multilayer Perceptron (MLP) and Long Short-Term Memory (LSTM) models, and it outperformed by achieving an overall accuracy of 94.23% and an Area under the Receiver Operating Characteristic (AUC-ROC) of 94%. The resulting GWP map was classified into three zones: high (74.98%), moderate (17.13%), and low (7.88%) potential. Validation using groundwater level data from twenty-nine monitoring wells yielded an accuracy of 77%. The findings demonstrate the effectiveness of CNN for GWP mapping and provide valuable insights for sustainable groundwater resource management, policy and decision-making.

Graphical Abstract