Multiple Data Imputation Methods Advance Risk Analysis and Treatability of Co-occurring Inorganic Chemicals in...

Area of Science:

• Environmental Science
• Hydrogeology
• Data Science

Background:

• Groundwater quality data is often sparse, hindering accurate risk assessment and management of inorganic pollutants.
• Limited budgets for sampling and analysis create significant data gaps, impacting exposure evaluation and treatment effectiveness.

Purpose of the Study:

• To compare the effectiveness of AMELIA and MICE multiple data imputation techniques for filling gaps in sparse groundwater quality data.
• To assess the impact of data imputation on the evaluation of groundwater pollutant risks and the identification of potential health hazards.

Main Methods:

• Utilized and compared AMELIA and MICE algorithms for multiple data imputation on sparse groundwater quality datasets.
• Analyzed field data and imputed data to evaluate the prevalence of inorganic pollutants exceeding Maximum Contaminant Levels (MCLs).

Main Results:

• AMELIA demonstrated superior performance over MICE in handling missing values, with MICE producing more outliers due to value overestimation.
• Imputed data revealed a substantial increase (2-5 times) in sampling locations potentially exceeding health-based limits compared to field data.
• Imputed data identified a higher number of samples with multiple co-occurring chemicals surpassing health-based levels.

Conclusions:

• Data imputation, particularly using AMELIA, significantly enhances the understanding of groundwater pollutant risks by addressing data sparsity.
• This framework aids state agencies in prioritizing sampling locations and chemicals for monitoring, optimizing resource allocation for groundwater protection.
• The study provides a strategic approach to reduce uncertainty in groundwater data collection, maximizing the impact of limited funding.