Various countries and communities are defining strategic energy plans driven by concerns for climate change and security of energy supply. Energy models can support this decision-making process. The long-term planning horizon requires uncertainty to be accounted for. To do this, the uncertainty of input parameters needs to be quantified. Classical approaches are based on the calculation of probability distributions for the inputs. In the context of strategic energy planning, this is often limited by the scarce quantity and quality of available data. To overcome this limitation, we propose an application-driven method for uncertainty characterization, allowing the definition of ranges of variation for the uncertain parameters. To obtain a proof of concept, the method is applied to a representative mixed-integer linear programming national energy planning model in the context of a global sensitivity analysis (GSA) study. To deal with the large number of inputs, parameters are organized into different categories and uncertainty is characterized for one representative parameter per category. The obtained ranges serve as input to the GSA, which is performed in two stages to deal with the large problem size. The application of the method generates uncertainty ranges for typical parameters in energy planning models. Uncertainty ranges vary significantly for different parameters, from [-2%,2%] for electricity grid losses to [-47.3%, 89.9%] for the price of imported resources. The GSA results indicate that only few parameters are influential, that economic parameters (interest rates and price of imported resources) have the highest impact, and that it is crucial to avoid an arbitrary a priori exclusion of parameters from the analysis. Finally, we demonstrate that the obtained uncertainty characterization is relevant by comparing it with the assumption of equal levels of uncertainty for all input parameters, which results in a fundamentally different parameter ranking. (C) 2017 Elsevier Ltd. All rights reserved.