A parsimonious methodological framework for short-term forecasting of groundwater levels

Area of Science:

• Hydrology
• Data Science
• Environmental Modeling

Background:

• Groundwater is a critical resource for drought mitigation.
• Limited monitoring data hinders accurate forecasting of groundwater levels.
• Existing models often require extensive data, limiting their applicability.

Purpose of the Study:

• To propose and evaluate a novel, parsimonious integrated method for short-term groundwater level forecasting.
• To develop a method with low data requirements, operational ease, and applicability.
• To assess the performance of the proposed method using geostatistics, meteorological variables, and artificial neural networks.

Main Methods:

• Integrated approach combining geostatistics, optimal meteorological exogenous variables, and artificial neural networks (ANNs).
• Application of the method to the Campo de Montiel aquifer in Spain.
• Evaluation of different ANN models, including NAR (Non-linear Autoregressive) and NARX (Non-linear Autoregressive with Exogenous input), and Elman networks.
• Analysis of correlations between groundwater levels and precipitation, considering effective precipitation.

Main Results:

• The proposed method demonstrates effectiveness in short-term groundwater level forecasting.
• NARX and Elman networks using effective precipitation showed the best performance (21.6% and 29.4% of cases, respectively).
• Achieved a mean Root Mean Square Error (RMSE) of 1.14 m on the test set and forecasting errors ranging from 0.76 m to 1.05 m for up to 6 months ahead.

Conclusions:

• The developed integrated method is a viable and efficient tool for short-term groundwater level forecasting, especially in data-scarce regions.
• The approach offers operational advantages and is relatively easy to implement.
• The study highlights the importance of considering effective precipitation and utilizing ANNs for improved hydrological predictions.