Site-Specific Attenuation Factor Estimation for Arsenic in Vadose Zone: A Data-Driven Framework Incorporating Soil Properties and Wet-Dry Cycles

Arsenic (As) contamination threatens public health as it migrates from soil surface to groundwater through the vadose zone. The attenuation factor (AF), defined as the ratio of initial As concentration to maximum concentration reaching the groundwater, quantifies As retention in vadose zone. While the U.S. Environmental Protection Agency recommends a default AF value of 1, this approach overlooks site-specific attenuation capacity of soils, potentially overestimating contamination. To improve As risk assessment, tailored datasets of AF for As that consider local soil variability are essential. The transport of As in vadose zone is often modeled by the Mobile-Immobile Model (MIM), which effectively accounts for the mass transfer within the stagnant water regime in the vadose zone. However, the lack of site-specific datasets for MIM-based transport parameters and the overlook of wet-dry cycle effects hinder accurately applying attenuation factor to As risk assessments. This study aimed to: (1) develop regression models to predict MIM-based solute transport parameters and As remobilization under repeated wet-dry cycles using soil properties, and (2) integrate these models into a comprehensive framework to estimate AF values for soils in South Korea.

First, we compiled 129 published data points, covering diverse soil textures, bulk densities, and MIM-based solute transport parameters such as mobile water content, dispersivity, and mass transfer coefficients. This dataset was used to train Random Forest (RF) regression models, where soil texture and bulk density served as input variables, and MIM-based solute transport parameters were the outputs. Second, we conducted experiments using 22 soil columns with varying organic matter content, iron content, particle size distribution, and bulk density to assess the influence of soil heterogeneity on As remobilization under repeated wet-dry cycles. Initial As concentrations in soil and As concentrations in leachate after the first wet-dry cycle were measured. A new parameter, Re, was introduced as the ratio of As concentration in leachate to the initial As concentration in soil. A separate RF model was developed to predict Re, using soil properties as input variables. Model performance was evaluated with the coefficient of determination (R²) to assess predictive accuracy. The outputs from these RF models, MIM-based solute transport parameters and Re were integrated to estimate site-specific AF values.

The RF models demonstrated excellent performance, with R² values exceeding 0.9, in predicting both MIM-based solute transport parameters and Re. Soil properties from 28 sites across South Korea were collected, encompassing diverse characteristics such as variations in texture, organic matter content, iron content, and vadose zone depths. Using the developed models, MIM-based solute transport parameters and Re values for these 28 sites were estimated. Subsequently, site-specific AF values were calculated, ranging from 4.26 to 26.07. This variability highlights the significant influence of heterogeneous soil properties on As attenuation in vadose zone. These findings underscore limitations of using the default AF value of 1 and validate the importance of site-specific analyses for accurate As risk assessments. The proposed methodology provides practical tools for estimating AF values for arsenic and enhancing risk assessments globally, particularly in areas vulnerable to As contamination.