Uncertainty analysis of the Carson River mercury transport model

Three computer models developed by the US Environmental Protection Agency (RIVMOD, WASP5 and MERC4) are used to simulate mercury transport and fate within the Carson River in west-central Nevada. While the transport of inorganic mercury is well understood and model prediction occurs with reasonable accuracy, it is recognized that the processes affecting methylmercury (MeHg) fate are more complicated. During January 1997 an extreme flood occurred which caused significant geomorphic change to the system. Following this flood, model error concerning MeHg water column concentrations increased substantially. This study addresses uncertainty in predicting water column MeHg in the hope of delineating change to MeHg transport caused by the 1997 flood. A preliminary sensitivity analysis shows that both the diffusion rate and the methylation/demethylation ratio (M/D) significantly affect MeHg water column concentrations, particularly during low discharge. It is discovered that the measured upper limit of M/D must be constrained from 0.26 to 0.065 to ensure accurate model calibration. A more comprehensive Monte Carlo analysis is conducted to test error propagation associated with uncertainty in methylation and demethylation rate coefficients. Methylation and demethylation rate assignments are assumed to vary independently with each parameter being characterized by a uniform distribution thereby reflecting user decision uncertainty. Monte Carlo results show that model uncertainty, defined by the 90% confidence interval (CI), in the lower portions of the Carson River converges upon 1.9 and 11.6 ng/l for steady discharges of 28.3 m3/s (1000 ft3/s) and 1.42 m3/s (50 ft3/s), respectively. The 90% CI encompasses the scatter of the observed data points, proving that it is not possible to suggest, with a high degree of statistical certainty, a significant change has occurred to MeHg transport and fate as a result of the 1997 flood.