Analytical Replication Improves Precision but Does Not Create Environmental Replication: Hierarchical Uncertainty and Decision Stability in Metal(loid) Soil Exposure Screening

Environmental exposure screening relies on concentration inputs shaped by hierarchical sampling and analytical design. This study investigated how analytical-unit selection and the allocation of replication across a nested soil-sampling hierarchy affect concentration precision, environmental inference, and downstream exposure-screening decisions. We reanalysed a published Leicester/Leicestershire topsoil dataset comprising 26 independent site-level composites, each represented by two physical laboratory samples, with 102 valid analytical determinations overall. Log-scale mixed models quantified variance components for 10 fully detected elements; 500 censor-compatible realisations addressed Cd and Cr; and 10,000 simulations compared reduced analytical designs with the full observed design (R4). Between-composite differences accounted for 65.6-93.8% of log-scale variability, with sensitivity estimates of 87.2% for Cd and 92.3% for Cr. At comparable analytical effort, one determination from each of two physical samples (R3) improved agreement with R4 relative to duplicating one physical sample (R2): median element MdAPE was 1.32% versus 4.27%, mean agreement within ±10% was 96.7% versus 77.5%, and mean rank displacement was 0.47 versus 1.42 positions. Treating the 102 analytical records as independent yielded five FDR-significant urban-rural signals, whereas analyses respecting the 26 independent environmental units yielded none after FDR adjustment. No hazard-index or cancer-risk-band classification switched relative to R4; mean Pb screening-classification switch probabilities were 9.57%, 7.72%, and 0.96% under R1, R2, and R3, respectively. Analytical replication improves measurement precision but does not create environmental replication.