Recent contributions to the ecological literature have questioned the continued usefulness of the classical model calibration paradigm in estimating parameters in coupled ecohydrological models. Schymanski (2007) and Schymanski et al. (2007, 2008) have demonstrated that the assumption of vegetation optimality precludes the need for site-specific data for estimating vegetation properties, transpiration fluxes, and CO2 assimilation. The goal of this article is twofold. We first show that significant advances in optimality-based vegetation modelling can be made if we embrace a novel concept of stochastic optimization that includes explicit recognition of parameter uncertainty. We adapted the original Vegetation Optimality Model (VOM) to a multi-layer soil and canopy vegetation optimality model, VOMmlsc with dynamically varying throughfall fraction. The DiffeRential Evolution Adaptive Metropolis (DREAM) algorithm is used to find parameter values with high values of net carbon profit (NCP), a proxy for biological fitness. We then show that significant variability exists in optimized vegetation properties and primarily transpiration fluxes from optimality of NCP. Seemingly, a myriad of vegetation species is possible that results in optimal values of NCP. Using data from a Douglas-fir plantation in The Netherlands, we found relative poor correspondence between modelled and measured ET and CO2-fluxes. The fitting of these two fluxes and values of the model parameters can be much improved when VOMmlsc is calibrated directly against these respective observations. Yet, the NCP values derived this way deviate considerably from their maximum possible value. This challenges the appropriateness of current weights to aggregate the various carbon costs and benefits into a single NCP scalar. Copyright © 2010 John Wiley & Sons, Ltd.