Prediction uncertainty of plume characteristics derived from a small number of measuring points

A small number of measuring points may inflict a bias on the characterisation of flow and transport based on field experiments in the unsaturated zone. Simulation of pure advective transport of a Gaussian plume through a setup of 30 regularly placed measuring points revealed regular temporal fluctuations about the real spatial moments. An irregular setup predicted both irregular fluctuations and larger discrepancies from the real value. From these considerations, a regular setup is recommended. Spatial moments were sensitive to the plume size relative to the distance between individual measuring points. To reduce prediction errors of the variance, the distance between the measuring points should be less than twice the standard deviation of the examined plume. The total size of the setup should cover several standard deviations of the plume to avoid mass being lost from the monitored area. Numerical simulations of a dispersing plume (comparing calculations based on 9000 nodes with 30 measuring points) revealed that vertical and horizontal centres of mass were predicted well at all degrees of heterogeneity, and the same was the case for horizontal variances. Vertical variances were more susceptible to prediction errors, but estimates were of the same order of magnitude as the real values. Received, October 1998/Revised, June 1999, October 1999/Accepted, November 1999