TO WHAT EXTENT DOES THE DEGREE OF MEANINGFUL MODEL COMPLEXITY DEPEND ON OBSERVATION DATA?

We systematically quantify to what extent vadose zone parameters can be constrained by including different types of observation data in the calibration process. Observation data considered are hydraulic heads, soil saturation, evaporation and transpiration. Besides assessing parameter uncertainty in relation to the availability of these observations, we quantify the uncertainty related to a prediction of a water table at a given time. Our analysis is based on the simulation of infiltration events through partially unsaturated 1d columns, as well as linear and non-linear methods to assess uncertainty. The results show that observations of the hydraulic head allow a significant reduction of predictive uncertainty. However, this it is not because any parameter is well estimated. Rather, this reduction is acquired knowledge that is applicable to combinations of parameters that contribute to a significant reduction of predictive uncertainty. This gives raise to interesting questions concerning the level of complexity required. The capacity of a recharge model to reduce the uncertainty of future groundwater level predictions does not rely on accurate estimation of all hydraulic properties affecting unsaturated zone water movement. Instead it relies on good estimation of only certain combinations of these properties. The key feature of the calibration process in this case is that it allows the model to make the predictions required, even though this ability pertains to combinations of parameters and not to individual parameters. Consequently, parameters and parameter combinations that are left unconstrained by that process are dispensable. It follows that the model may lose little if it were simplified in ways suggested by the calibration process. This point is illustrated by comparing predictions of groundwater recharge obtained through a lumped parameter model and a fully saturated/ unsaturated model.