Equifinality, sensitivity and predictive uncertainty in the estimation of critical loads

The impacts of acidifying atmospheric deposition to soil and water resource s are commonly calculated utilising predictive mathematical models. The est imation of the predictive uncertainty inherent in these models is important since the model predictions are increasingly being used as a scientific ba sis for decisions on emission abatement policies and strategies in Europe. When predictive uncertainty is taken into account it may significantly affe ct conclusions ascertained from model predictions. The Generalised Likeliho od Uncertainty Estimation (GLUE) approach is used here in the estimation of predictive uncertainty of PROFILE, a steady-state biogeochemical model. GL UE is based on Monte Carlo simulation and recognises the possible equifinal ity of parameter sets. With this methodology it is possible to make an asse ssment of the likelihood of a parameter set being an acceptable simulator o f a system when model predictions are compared to measured field data. The GLUE methodology is applied to PROFILE simulations of five European researc h sites. The results have revealed that the model is unable to reproduce th e characteristics of soil water chemistry consistently, and that the result ing predicted critical loads must be associated with significant uncertaint y. The study also demonstrates that a wide range of parameter sets exist th at give acceptable simulations of site characteristics as well as a broad r ange of critical load values that are consistent with the site data. A sens itivity analysis is performed for simulations of data sets from each site; this is employed to evaluate the role of the model parameters in forcing th e predictions. Results of the sensitivity analyses show that, in general, s ite predicted soil chemistry is driven by atmospheric inputs and mineral we athering rates are determined by soil physical properties. (C) 1999 Elsevie r Science B.V. All rights reserved.