Benefits and Cautions in Data Assimilation Strategies: An Example of Modeling Groundwater Recharge

Area of Science:

• Hydrology
• Hydrogeology
• Environmental Science

Background:

• Groundwater models are crucial for assessing water resources.
• Real-time groundwater assessments benefit from incorporating recent observational data.
• Estimating time-varying recharge in fractured-rock aquifers presents challenges.

Purpose of the Study:

• To compare the efficacy of different data assimilation strategies in improving groundwater recharge models.
• To evaluate forecasting, filtering, and fixed-lag smoothing (FLS) methods for estimating aquifer states and recharge.
• To assess the trade-offs between model accuracy, uncertainty reduction, and mass conservation.

Main Methods:

• Applied the Kalman Filter to a data-driven, mechanistic recharge model.
• Implemented and compared three data assimilation approaches: forecasting, filtering, and fixed-lag smoothing (FLS).
• Analyzed time-varying water-table altitude (h) and recharge estimates, including their error covariances.

Main Results:

• Forecasting produced high error covariance, indicating significant uncertainty.
• Filtering and FLS, by assimilating recent observations, yielded recharge estimates that better matched water-table dynamics and reduced uncertainty compared to forecasting.
• While filtering and FLS reduced uncertainty, they did not guarantee mass conservation, unlike forecasting.

Conclusions:

• Data assimilation strategies, particularly filtering and FLS, significantly improve real-time groundwater condition estimates by reducing uncertainty.
• The choice between data assimilation methods requires balancing improved state estimation against potential loss of mass conservation.
• Findings are applicable to various groundwater process models sensitive to system inputs and external data.