The present research analyses the error on prediction obtained under different data availability scenarios to determine which measurements contribute to an improvement of model prognosis and which not. A fully coupled 2D hydromechanical model of a water retaining dam is taken as an example. Here, the mean effective stress in the porous skeleton is reduced due to an increase in pore water pressure under drawdown conditions. Relevant model parameters are ranked by scaled sensitivities, Particle Swarm Optimization is applied to determine the optimal parameter values and model validation is performed to determine the magnitude of error forecast. We compare the predictions of the optimized models with results from a forward run of the reference model to obtain actual prediction errors. The analyses presented here were performed to 31 data sets of 100 observations of varying data types. Calibrating with multiple information types instead of only one sort, brings better calibration results and improvement in model prognosis. However, when using several types of information the number of observations have to be increased to be able to cover a representative part of the model domain; otherwise a compromise between data availability and domain coverage prove best. Which type of information for calibration contributes to the best prognoses, could not be determined in advance. For the error in model prognosis does not depends on the error in calibration, but on the parameter error, which unfortunately can not be determined in reality since we do not know its real value. Excellent calibration fits with parameters’ values near the limits of reasonable physical values, provided the highest prognosis errors. While models which included excess pore pressure values for calibration provided the best prognosis, independent of the calibration fit.