This paper outlines the steps followed to construct hydro-climatic projections of basin-scale runoff and their associated uncertainties over the Quebec/Labrador peninsula. First, we show that the physically-based Canadian RCM (Regional Climate Model) is able to reproduce basin-scale annual runoff within observational errors. The robustness of the CRCM at simulating annual runoff at the basin scale is studied through an analysis of the models intrinsic internal noise (i.e., internal variability related to the non-deterministic nature of the climate system), which we find to be small (with respect to observational errors and the interannuol variability of observed runoff). The CRCM's main advantage is that it is constructed with balanced land and atmosphere water and energy budgets, and includes feedbacks between the surface and the atmosphere; providing variables that are all internally consistent. However, due to weaknesses in the representation of snow and land-surface processes in CLASS 2.7, the simulated intra-annual runoff is not adequately reproduced. Sensitivity experiments show that domain size has an important effect on simulated annual runoff and that surface scheme and driving reanalyses have less influence but still remain significant. These findings imply that not only should the experimental configuration of a simulation be carefully defined according to the area of interest, but also that one must consider results from more than a single RCM simulation. Following these basic steps, more trustworthy climate change data can be provided to water resource managers. Through the provision of an ensemble of regional climate projections, it is then possible to evaluate the climate change signal and the associated level of confidence.
