304029

Calibration and Reliability in Groundwater Modelling: From Uncertainty to Decision Making (Proceedings of ModelCARE’2005, The Hague, The Netherlands, June 2005). IAHS Publ. 304, 2006. pp.202–208.

Comparison of stochastic and regression based methods for quantification of predictive uncertainty of model-simulated wellhead protection zones in heterogeneous aquifers

STEEN CHRISTENSEN¹, CATHERINE MOORE² & JOHN DOHERTY²

1 Department of Earth Sciences, University of Aarhus, Ny Munkegade b. 520,
DK-8000 Aarhus C, Denmark

sc@geo.au.dk

2 School of Engineering, University of Queensland, Brisbane, Queensland, Australia

Abstract For a synthetic case we computed three types of individual prediction intervals for the location of the aquifer entry point of a particle that moves through a heterogeneous aquifer and ends up in a pumping well.

(a) The nonlinear regression-based interval (Cooley, 2004) was found to be nearly accurate and required a few hundred model calls to be computed.

(b) The linearized regression-based interval (Cooley, 2004) required just over a hundred model calls and also appeared to be nearly correct.

(c) The calibration-constrained Monte Carlo interval (Doherty, 2003) was found to be narrower than the regression-based intervals but required about half a million model calls. It is unclear whether or not this type of prediction interval is accurate.

Keywords accuracy; computational requirements; Monte Carlo method; prediction interval; predictive uncertainty; regression based method; wellhead protection zone