Data and Reproducibility Package for "Stress-Testing Organic-Carbon Normalization: Chemical-Domain and Environmental-Window Dependence of Sorption-Based Mobility Assessment"

Abstract

This repository contains the curated data, derived analysis datasets, modelling outputs, validation results, figure-ready data, and reproducibility materials supporting the study “Stress-Testing Organic-Carbon Normalization: Chemical-Domain and Environmental-Window Dependence of Sorption-Based Mobility Assessment.” The study evaluates when organic-carbon normalization of solid–water sorption coefficients provides transferable information for contaminant mobility assessment. Normalization performance is quantified using \(R_{\mathrm{norm}} = \mathrm{Var}(\log K_{oc}) / \mathrm{Var}(\log K_d)\), where \(R_{\mathrm{norm}}<1\) indicates that normalization reduces variability relative to the unnormalized \(K_d\), whereas \(R_{\mathrm{norm}}>1\) indicates increased variability. The associated organic-carbon response coefficient, \(\beta_{OC}\), is used to provide a mechanistic interpretation of normalization performance. The repository covers three chemical domains: PFAS, pharmaceuticals, and pesticides. The PFAS dataset was used to establish the primary analytical and predictive framework and to evaluate chemical-, study-, and double-out-of-domain generalization. Frozen PFAS-derived relationships were subsequently tested in two independent external domains: pharmaceuticals and a pesticide dataset based on the Motoki soil-sorption study. The final cross-domain synthesis contains 109 chemical-level \(R_{\mathrm{norm}}\) targets comprising 38 PFAS, 54 pharmaceuticals, and 17 pesticides. The analyses include chemical and environmental feature representations, structural applicability-domain assessment, environmental-domain support, uncertainty analysis, external transfer evaluation, fixed-\(K_{oc}\) threshold-instability assessment, and sensitivity of normalization performance to the environmental organic-carbon window. Key repository components include: curated and processed chemical-level and measurement-level datasets; \(R_{\mathrm{norm}}\), \(\log_{10}(R_{\mathrm{norm}})\), and \(\beta_{OC}\) targets; molecular descriptor and structural-similarity outputs; environmental-context variables and applicability-domain results; frozen PFAS model-validation and external-transfer outputs; pharmaceutical and pesticide external-domain analyses; Motoki pesticide censoring, complete-case, leave-one-soil-out, and organic-carbon-window sensitivity outputs; \(K_{oc}\) threshold-instability results; cross-domain synthesis tables; figure-ready datasets and publication figures; analysis scripts and provenance/QC summaries required to reproduce the reported results. No additional exploratory modelling beyond the analyses reported in the associated manuscript is included. The repository is intended to provide transparent provenance, reproducibility, and machine-readable access to the data underlying the study. Associated manuscript:Stress-Testing Organic-Carbon Normalization: Chemical-Domain and Environmental-Window Dependence of Sorption-Based Mobility AssessmentTarget journal: Environmental Science & Technology. Note: The three chemical domains originate from different study designs and environmental sampling frameworks; cross-domain comparisons should therefore be interpreted as assessments of normalization behavior and transferability rather than as measurements from an identical experimental population.