Estimating Uncertainty in Long-Term Reconstructions

Summary

A “regionalization” methodology is presented that uses a conceptual model of long-term water quality responses to acidic deposition in a Monte Carlo framework to simulate the observed distributions of water quality variables derived from regional water quality surveys. The regional model contains physical parameters that describe the hydrogeochemical characteristics of the region and deposition parameters which describe the magnitude and long-term temporal sequence of acidic deposition. The methodology has been applied to data from a survey of lakes in southern Norway. Sensitivity analyses of the regionally calibrated model suggest that the physical parameters are well determined and that the model is not over-parameterized. We examine here the additional effects of uncertainties in the deposition parameters. Monte Carlo procedures are used to generate an ensemble of long-term historical deposition sequences in which the timing and magnitude of increases of acidic deposition are varied. The effects of this deposition uncertainty on simulation of water quality distributions are of the same order as the effects produced by uncertainty in physical parameters. The estimated uncertainties are non-stationary.