Models should account for correlated observation errors



Scientists need to account for observation errors in models. In some cases, such as when multiple observations are derived from a single direct measurement, the errors in different observations are correlated. Noting that few groundwater studies have taken into account correlated observational errors, Tiedeman and Green investigated the effect of error correlation on parameter estimates, predictions, and measures of uncertainty in a simple generic inverse model and in a more complex groundwater transport model of denitrification. They compared model results obtained with and without error correlations and found that not accounting for error correlations can significantly affect the results.