This work illustrates the application to a simulated solar collector facility of a recently developed, comprehensive, predictive modeling methodology for obtaining optimally predicted best-estimate results, with reduced uncertainties. The application of the very efficient adjoint sensitivity analysis methodology (ASAM) for nonlinear systems is also illustrated by computing exactly the first-order sensitivities of selected facility responses to all model parameters. These sensitivities are used to rank the importance of parameters in contributing to response uncertainties, and also serve within the predictive methodology as the weighting functions for propagating uncertainties of the model parameters and for assimilating measurements and simulations. The results produced by the predictive modeling procedure are optimally predicted values for the responses and for all model parameters, with reduced predicted uncertainties that are smaller than either the measured or the computed uncertainties. The amount of reduction is controlled by the magnitude of the respective sensitivities: the larger the magnitude of the sensitivities, the larger the reduction in the predicted uncertainties. The predictive methodology presented in this work can be used for validating simulation models, and for designing and/or improving the performance of experimental installations. Current limitations of this predictive modeling methodology are also highlighted, along with ongoing work towards generalizing and significantly extending its applicability. (C) 2015 Elsevier Ltd. All rights reserved.