|
Determining the best observation wells to predict groundwater depth using ANFIS, based on different training algorithms
|
Abbas Sedghamiz *   |
| Shiraz university |
|
|
Abstract: (122 Views) |
| The excessive increase in cultivated land for maximizing food production or boosting agricultural exports—especially in arid regions—alongside factors such as climate change has led to immense pressure on groundwater resources and a significant decline in groundwater levels in many of the country’s aquifers. Therefore, to effectively and optimally manage these resources, it is essential to make accurate forecasts based on all prevailing conditions in these aquifers, particularly the precise fluctuations in groundwater depth and level. The aim of this study, conducted in the Qotbabad region of Jahrom County in Fars Province, is to identify the wells that provide the most accurate predictions of groundwater depth in other wells. For this purpose, the Adaptive Neuro-Fuzzy Inference System (ANFIS) was used along with different training algorithms (Hybrid, GA, PSO). To evaluate the models' accuracy, indices such as Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Mean Absolute Error (MAE) were employed. Based on the results, Observation Well No. 2 was identified as the strongest predictor respectively for Observation Wells No. 1 and 2, while Observation Well No. 5 was the weakest predictor for Wells No. 3 and 6. Furthermore, Observation Well No. 4, classified as a medium-level predictor, emerged as the best predictor for Wells No. 2, 5, and 7, and the second-best predictor for Wells No. 1, 3, and 6. Therefore, unlike the other wells, Well No. 4 consistently ranked first or second in estimating groundwater depth in other observation wells. Among these, the strongest linear relationship was found for Observation Well No. 4. Its best predictor was Observation Well No. 2, with an average coefficient of determination (R²) of 0.9945 across the three training methods (Hybrid, GA, and PSO). On the other hand, the weakest linear relationship was observed for Observation Well No. 3, whose best predictor—Observation Well No. 5—produced an average R² of 0.7435 using the three mentioned training algorithms |
Article number: 7 |
| Keywords: Groundwater, Adaptive Neouro- Fuzzy Inference System, Training Algorithme, precies Index |
|
|
|
|
Type of Study: Research |
Subject:
Special
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
