Modelling uncertainty in Critical Load Models
A Monte Carlo approach was used together with Critical Load models to assess parameter sensitivity and model uncertainty. An extensive programme of testing uncertainties in the acidity critical load model used for terrestrial ecosystems was carried out at the Liphook site in Hampshire. The sensitivities to choice of input parameters, their statistical distribution type and their intercorrelation were calculated and the methods applied to other coniferous forest sites. It was hoped that this would identify parameters which could be selectively targeted for research to narrow output uncertainties, but almost any input parameter proved potentially important. The methods were also applied to heathland, unmanaged woodland, and critical loads for nutrient nitrogen, covering the entire range of terrestrial critical load methods used in the UK.
An uncertainty analysis of acidity critical loads for freshwaters was carried out using similar methods, and applied to the 22 sites of the UK Acid Waters Monitoring Network. Here one parameter, the current base cation concentration in waters, dominated the contribution of input parameters to total uncertainty. For both aquatic and terrestrial ecosystems, output parameter uncertainty was generally considerably less than input parameter uncertainty, thought to be due to a 'compensation of errors' mechanism. The methodology was then applied on a national scale using managed coniferous forests as an example. Similar patterns of sensitivity were observed, but also variability between regions.
Back to the Critical Load study home page
Back to research home page
