ARCHI: A New R Package for Automated Imputation of Regionally Correlated Hydrologic Records

Area of Science:

• Hydrology
• Data Science
• Environmental Science

Background:

• Missing data in hydrological records poses significant challenges for resource assessment, process understanding, and predictive modeling.
• Accurate imputation of missing data is crucial for reliable hydrological analysis.

Purpose of the Study:

• To introduce ARCHI (Automated Regional Correlation Analysis for Hydrologic Record Imputation), a novel open-source R package.
• To provide a comprehensive tool for aggregating, imputing, clustering, and visualizing regionally correlated hydrologic records.

Main Methods:

• ARCHI utilizes linear regression, employing complete "reference" records to impute missing data in "target" records.
• An iterative algorithm allows sites to serve as both target and reference, expanding the pool of complete data.
• Geospatial, correlation, and group-based filters, along with model-acceptance criteria, mitigate spurious correlations.

Main Results:

• Benchmarking on groundwater-level datasets from California and New York demonstrates ARCHI's computational efficiency and imputation accuracy.
• ARCHI offers model transparency and is evaluated against widely used multivariate imputation software.

Conclusions:

• ARCHI provides an accessible and intuitive solution for imputing missing hydrologic data, enhancing the reliability of hydrological studies.
• The package's GUI and visualization tools facilitate interactive parameterization and result interpretation.