Time Series Analysis of Nonlinear Head Dynamics Using Synthetic Data Generated with a Variably Saturated Model.

Area of Science:

• Hydrogeology
• Environmental Modeling
• Time Series Analysis

Background:

• Groundwater head fluctuations are influenced by complex hydrological processes, including precipitation and evaporation.
• Accurate simulation of these dynamics is crucial for effective water resource management.
• Existing time series models may not fully capture the nonlinear behavior of groundwater systems.

Purpose of the Study:

• To assess the performance of linear and nonlinear time series models in simulating synthetic groundwater head data.
• To compare the accuracy of these models against a numerical Richards' equation model.
• To provide tools for evaluating data-driven hydrological models.

Main Methods:

• Generated synthetic groundwater head series using a numerical model solving Richards' equation for variably saturated flow.
• Simulated head responses to precipitation and evaporation under different soil types and unsaturated zone thicknesses.
• Applied and evaluated both linear and nonlinear time series models using R-squared values.

Main Results:

• Linear time series models achieved R-squared values from 0.67 to 0.96.
• Nonlinear time series models, incorporating a root zone reservoir, consistently achieved R-squared values above 0.9.
• The nonlinear model's precipitation event response closely matched the numerical model's output.

Conclusions:

• Nonlinear time series models demonstrate superior performance in simulating groundwater head dynamics compared to linear models.
• The developed synthetic data generation scripts can be valuable for testing various data-driven hydrological models.
• Accurate simulation of groundwater recharge and head response is achievable with advanced time series techniques.