Using semivariogram parameter uncertainty in hydrogeological applications.

Area of Science:

• Geosciences
• Hydrogeology
• Spatial Statistics

Background:

• Geostatistical methods like kriging and simulation are vital in groundwater hydrology for spatial interpolation and risk assessment.
• These methods rely on spatial variability models (semivariograms) inferred from limited data, introducing uncertainty.
• Using a single fitted model with small datasets can lead to biased predictions in groundwater resource management.

Purpose of the Study:

• To address the uncertainty associated with semivariogram model selection in geostatistics.
• To develop a methodology that accounts for the range of plausible semivariogram models.
• To improve the reliability of geostatistical predictions in groundwater hydrology.

Main Methods:

• Maximum likelihood inference to quantify semivariogram model uncertainty.
• Propagating uncertainty from plausible semivariogram models into kriging and simulation applications.
• Applying the methodology to predict groundwater head using log-transmissivity data.

Main Results:

• Demonstrated a method to evaluate and incorporate semivariogram parameter uncertainty.
• Showcased the impact of using a range of plausible models versus a single model on final geostatistical results.
• Successfully applied the approach in a case study of the Vega de Granada aquifer.

Conclusions:

• Accounting for semivariogram uncertainty is crucial for robust geostatistical analysis in groundwater hydrology.
• The proposed method enhances the reliability of spatial interpolation and risk assessment by considering model plausibility.
• This approach provides a more realistic representation of prediction uncertainty in groundwater studies.