Delineation of Hazardous Areas and Additional Sampling Strategy in Presence of a Location-Specific Threshold

Abstract

This paper presents a methodology to account for multiple sources of uncertainty (measurement and interpolation errors for pollutant concentrations, spatially variable regulatory thresholds) in the mapping of the probability of contamination. The approach, which involves kriging of soft indicator data and propagation of uncertainty by Latin-Hypercube Sampling (LHS) of ccdfs, is applied to a 216 km(2) airborne Cd-contaminated area in Belgium. The aim is to predict the probability of exceeding a regulatory sanitation threshold (ST) for vegetable gardens which depends on both soil organic matter (OM) and clay content. A validation study shows that the ccdfs provide accurate models of the uncertainty about soil properties. The coefficient of variation of the ccdf of the difference D=ST-[Cd] is proposed as a criterion to decide where additional samples should be collected. In the case study, this criterion allows one to identify mostly locations that were wrongly declared safe, while large prediction errors were detected by sampling locations where the variance of the Cd ccdf is high.