Performance Analysis of ANN Prediction for Groundwater Level Considering Regional-Specific Influence Components

Area of Science:

• Hydrology
• Hydrogeology
• Artificial Intelligence

Background:

• Groundwater level (GWL) is influenced by multiple dynamic factors, including precipitation, river stage (RS), sea level, and dewatering.
• Accurate prediction of GWL is crucial for water resource management and infrastructure planning.

Purpose of the Study:

• To investigate the influence of various components on groundwater level (GWL) prediction using artificial neural networks (ANN).
• To identify site-specific factors that enhance the reliability of ANN-based GWL predictions across diverse hydrologic and geologic settings.

Main Methods:

• Collected data from six regions with varying hydrologic and geologic conditions.
• Employed artificial neural networks (ANN) with different input combinations to predict GWL.
• Analyzed the correlation between GWL and influencing components, including river stage (RS) and precipitation moving average (MA).

Main Results:

• In urban areas, river stage (RS) was identified as the primary influence on GWL.
• In rural areas, ground permeability significantly impacted GWL, with the moving average (MA) of precipitation proving effective.
• The inclusion of RS and MA in ANN inputs significantly improved prediction performance in urban and rural areas, respectively.

Conclusions:

• Site-specific influence components must be identified and incorporated into ANN models for enhanced and reliable GWL prediction.
• The impact of learning data length (LDL) on prediction accuracy was found to be less significant compared to the identified influence components.