PyLEnM

PyLEnM provides machine-learning methods for long-term groundwater contamination monitoring, including data QA/QC, spatiotemporal data integration, clustering, proxy-based spatial interpolation, automated model selection and parameter tuning, and well-network optimization.

Key Features:

• Data Management Utilities: Implements quality assurance and quality control (QA/QC) and identifies coincident/colocated spatial and temporal datasets to ensure reliable data integration.
• Data Ingestion and Processing: Automates ingestion and processing of publicly available spatial data layers for use as predictors in analyses.
• Time Series/Multianalyte Clustering: Performs time-series and multianalyte clustering to group wells with similar groundwater dynamics for interpolation and monitoring optimization.
• Automated Model Selection and Parameter Tuning: Compares multiple regression models and tunes parameters to select models for spatial interpolation tasks.
• Proxy-Based Spatial Interpolation: Uses spatial data layers or in situ measurable variables as predictors for contaminant concentrations and groundwater levels.
• Well Optimization Algorithm: Identifies optimal subsets of wells to maintain robust spatial interpolation capabilities over extended monitoring periods.

Scientific Applications:

• Groundwater Monitoring: Supports long-term monitoring of contaminant concentrations and groundwater levels in environmental science and public health contexts.
• Well Network Design: Informs selection and optimization of monitoring well placements to preserve interpolation accuracy over time.
• Spatial Prediction for Remediation: Enhances spatial interpolation models used to predict contaminant spread and concentration levels for environmental management and remediation planning.

Methodology:

Applies QA/QC, coincident/colocated data identification, automated ingestion of spatial data layers, time-series and multianalyte clustering, regression model comparison and parameter tuning, proxy-based spatial interpolation, and a well-optimization algorithm to field datasets such as the Savannah River Site F-Area.