Groundwater Potential Mapping Using an Integrated Ensemble of Three Bivariate Statistical Models with Random Forest and Logistic Model Tree Models
| Autor: | Soo-Mi Choi, Abolghasem Sadeghi-Niaraki, S. Vahid Razavi-Termeh |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2019 |
| Předmět: |
lcsh:TD201-500
lcsh:Hydraulic engineering Receiver operating characteristic Geography Planning and Development Frequency ratio hybrid model Statistical model Bivariate analysis data mining models Aquatic Science GIS Biochemistry Logistic model tree Random forest bivariate statistic models lcsh:Water supply for domestic and industrial purposes lcsh:TC1-978 Statistics Environmental science groundwater potential mapping (GPM) Groundwater Statistic Water Science and Technology |
| Zdroj: | Water, Vol 11, Iss 8, p 1596 (2019) Water Volume 11 Issue 8 |
| ISSN: | 2073-4441 |
| Popis: | In the future, groundwater will be the major source of water for agriculture, drinking and food production as a result of global climate change. With increasing population growth, demand for groundwater has increased. Therefore, sustainable groundwater storage management has become a major challenge. This study introduces a new ensemble data mining approach with bivariate statistical models, using FR (frequency ratio), CF (certainty factor), EBF (evidential belief function), RF (random forest) and LMT (logistic model tree) to prepare a groundwater potential map (GPM) for the Booshehr plain. In the first step, 339 wells were chosen and randomly split into two groups with groundwater yields above 11 m3/h. A total of 238 wells (70%) were used for model training, and 101 wells (30%) were used for model validation. Then, 15 effective factors, including topographic and hydrologic factors, were selected for the modeling. The accuracy of the groundwater potential maps was determined using the ROC (receiver operating characteristic) curve and the AUC (area under the curve). The results show that the AUC obtained using the CF-RF, EBF-RF, FR-RF, CF-LMT, EBF-LMT and FR-LMT methods were 0.927, 0.924, 0.917, 0.906, 0.885 and 0.83, respectively. Therefore, it can be inferred that the ensemble of bivariate statistic and data mining models can improve the effectiveness of the methods in developing a groundwater potential map. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|