Predicting Groundwater Levels in Ogallala Aquifer Wells Using Hierarchical Cluster Analysis and Artificial Neural Networks

The Ogallala Aquifer, located in the Central Plains of the United States, is essential for agricultural irrigation and public water supply. Indiscriminate pumping from the aquifer has caused several negative impacts, such as deterioration of water quality and depletion of groundwater levels, which urgently demand better management. This paper applies hierarchical cluster analysis (HCA) and artificial neural networks (ANNs) for predicting annual groundwater levels in 403 wells of the Ogallala Aquifer. First, the methodology employed HCA to cluster homogeneous wells based on the time series of groundwater levels. Then, the study calibrated an ANN model for each cluster (composed of one or more wells) using previous annual values of groundwater levels as input. The HCA results showed a particular pattern in the spatial distribution of the 30 found clusters, revealing that the Ogallala Aquifer holds higher groundwater levels in the western part, which gradually decrease, advancing to the east. The ANN models provided proper predictions even for wells outside of the calibration data set. This investigation concludes that the integration of HCA and ANN enabled single models to accurately forecast annual groundwater levels for sets of wells in the Ogallala Aquifer.

Automated summary

This study uses hierarchical cluster analysis (HCA) and artificial neural networks (ANNs) to predict annual groundwater levels in the Ogallala Aquifer. The results show that the spatial distribution of groundwater levels in the aquifer follows a particular pattern, with higher levels in the west and gradually decreasing levels towards the east. The integration of HCA and ANN enables accurate predictions for sets of wells in the aquifer.