Validation of handheld X-ray fluorescence against wavelength-dispersive X-ray fluorescence for heavy metal quantification in contaminated soils

Abstract

Handheld X-ray fluorescence (hXRF) technology has revolutionized environmental site assessment by providing rapid, in-situ elemental analysis. Accuracy of these devices in complex, highly contaminated soil matrices remains questionable and validation against laboratory reference methods remains limited. This study evaluates the performance of an Olympus Vanta C Series handheld XRF operated in Alloy Plus mode against laboratory-based wavelength dispersive X-ray fluorescence (WDXRF) using soil samples from a ballistic contamination gradient at the Mirpur Cantonment firing range in Dhaka, Bangladesh. Nine soil samples were collected across a contamination gradient (three sites, three replicates per site). From the study, strong correlations for lead (r=0.972; R2 = 0.945) and iron (r = 0.800) were found with significant systematic bias. In this regard, lead concentrations were overestimated by hXRF about 275.7% with a mean bias = +12.11%, likely due to limited sample size, matrix effects and spectral interferences. Site-specific linear regression correction models were developed and validated using Leave-One-Out Cross-Validation (LOOCV) to mitigate these biases.  Empirical correction factors resulted in an RMSE for lead that was reduced from 16.82% to 2.72%, representing an improvement in accuracy of 83.8%. This study concludes that empirical calibration is necessary for quantitative risk assessment, even if uncorrected hXRF data is appropriate for delineating contamination boundaries and relative screening. A tiered decision framework, as proposed, should guide environmental practitioners on the integration of hXRF screening with confirmatory laboratory analysis.