Scholarlink

<p>The growing water demand related to well-being of the populations, including community of remote areas and the spreading of highly water-demanding agriculture practices are two causes of the water stress in developing countries. They are consequently also stressing the finances the municipalities of developing country, including Cameroon, being in the frame of the decentralization processes the water supply for, strategic services such as health centers schools, and households in both villages and rural areas. The main objective of this study is to map the groundwater potential zones (GwPZs) of the study area based on 16 thematic layers elaborated mainly using satellite images (8), existing maps (2) and subsurface data, based on geoelectric survey (6). The six (6) maps of GwPZs have been elaborated using four bivariate statistical methods, namely Frequency Ratio (FR), Shannon Entropy (SE), Weight of Evidence (WofE) and Certain Factor (CF) and two combined statistical artificial neural network sets namely Multilayer Perceptron (MLP) coupled with FR (MLP-FR), and CF (MLP-CF). The GwPZs have been mapped by weighting and combining in ArcGIS environment the classes or categories distinguished for each of the sixteen thematic layers. The data management including the weights assigned to each class distinguished for the 16 selected thematic layers of have been according the adopted models and the reliability of the elaborate maps have been tested and validated using the area under the receiver operating characteristic curves (AUC). For each map, the GwPZs have been distinguished into four distinct classes: low, moderate, high and very high. Although the shape and extension of the zones with same class are different for the 6 elaborated maps the total extension of the area assigned to the different class are comparable. In particular the class Moderate (from 56.67 to 79.1%) and the class Very High (from 0.53 to 2%) shoe comparable results. The class Low (from 2 to 77%) and class High (from 9.1 to 34%) show instead the main differences. However, based on the AUC, the results show that the MLP-CF method has the highest success rate (AUC = 0.86), with respect to SE (0.82), MLP-FR (0.78), FR (0.75), CF (0.73) and WofE (0.72). And the FR model has the highest prediction accuracy (AUC = 0.76), with respect to WofE (0.74), CF (0.70), MLP-CF (0.69), SE (0.67) and MLP-FR (0.67). Therefore, the FR model is the most reliable for delimiting the GwPZs in the locality of Bafia.</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

Bivariate statistical and neural network models to map groundwater potential zones in Bafia area (Central Cameroon)

  • 期刊论文

  • 发表日期：

    2024年12月31日

查看全文
  • Anita Ngouokouo Tchikangoua,
  • Françoise Martine Enyegue A Nyam,
  • Serges Raoul Kouamou Njifen,
  • William Assastsé Teikeu,
  • Théophile Ndougsa Mbarga,
  • Nicola Perilli

摘要

The growing water demand related to well-being of the populations, including community of remote areas and the spreading of highly water-demanding agriculture practices are two causes of the water stress in developing countries. They are consequently also stressing the finances the municipalities of developing country, including Cameroon, being in the frame of the decentralization processes the water supply for, strategic services such as health centers schools, and households in both villages and rural areas. The main objective of this study is to map the groundwater potential zones (GwPZs) of the study area based on 16 thematic layers elaborated mainly using satellite images (8), existing maps (2) and subsurface data, based on geoelectric survey (6). The six (6) maps of GwPZs have been elaborated using four bivariate statistical methods, namely Frequency Ratio (FR), Shannon Entropy (SE), Weight of Evidence (WofE) and Certain Factor (CF) and two combined statistical artificial neural network sets namely Multilayer Perceptron (MLP) coupled with FR (MLP-FR), and CF (MLP-CF). The GwPZs have been mapped by weighting and combining in ArcGIS environment the classes or categories distinguished for each of the sixteen thematic layers. The data management including the weights assigned to each class distinguished for the 16 selected thematic layers of have been according the adopted models and the reliability of the elaborate maps have been tested and validated using the area under the receiver operating characteristic curves (AUC). For each map, the GwPZs have been distinguished into four distinct classes: low, moderate, high and very high. Although the shape and extension of the zones with same class are different for the 6 elaborated maps the total extension of the area assigned to the different class are comparable. In particular the class Moderate (from 56.67 to 79.1%) and the class Very High (from 0.53 to 2%) shoe comparable results. The class Low (from 2 to 77%) and class High (from 9.1 to 34%) show instead the main differences. However, based on the AUC, the results show that the MLP-CF method has the highest success rate (AUC = 0.86), with respect to SE (0.82), MLP-FR (0.78), FR (0.75), CF (0.73) and WofE (0.72). And the FR model has the highest prediction accuracy (AUC = 0.76), with respect to WofE (0.74), CF (0.70), MLP-CF (0.69), SE (0.67) and MLP-FR (0.67). Therefore, the FR model is the most reliable for delimiting the GwPZs in the locality of Bafia.