Uncertainty in a regional-scale assessment of cadmium accumulation in the Netherlands

Atmospheric deposition and agricultural activities may cause the cadmium (Cd) content in the rural environment in the Netherlands to exceed legal environmental standards. A simple process-oriented model, SOACAS, in combination with a Geographical Information System was used to assess the magnitude of the Cd accumulation on a regional scale. The objective of this study was to quantify the uncertainty in the Cd accumulation due to data uncertainty, Another objective was to study whether maps of simulated and observed Cd contents were statistically different when the uncertainty in both maps was considered. From a Monte Carlo analysis we derived that the model behaved virtually linear within the range of model inputs considered, and concluded that first-order uncertainty analysis (FOUA) was appropriate for mapping prediction uncertainties. The maps crested by FOUA indicated that the contribution of individual model parameters to the total uncertainty was soil dependent, and that the pedotransfer function for Cd sorption gave the largest contribution to the total uncertainty. In an earlier analysis we found that in 25% of the total area SOACAS underestimated the average levels of current Cd contents by >50%. The model predictions and the observations were statistically different at a much smaller area (10% of the total area), showing that ignoring uncertainty may result in misleading interpretations when the model is compared with field measurements.