Perceptual catchment models are a key first step towards hydrologic synthesis and provide the process foundation of predictive models. However, their development is usually limited to the headwater catchment scale where field work can be done. Here we develop a perceptual model approach for a regional scale catchment. We use a systematic "top-down" approach based on an expert-driven interpretation of the streamflow responses of nested subcatchments. This approach complements the traditional "bottom-up" perceptual model development based on fieldwork observations, which is impractical at the regional scale due to measurement limitations. As an example to illustrate our approach and demonstrate its feasibility, we use the 27,100 km² Moselle catchment, with 26 gauged subcatchments. Our perceptual model of the Moselle catchment presents a description of dominant processes that we consider a-posteriori reasonable, but not a-priori obvious: where precipitation, much more than evaporation or groundwater exchange controls the spatial variability of average streamflow, lithology influences the partitioning between baseflow and quickflow, and topography and land use control hydrograph lag times. A striking feature of our analysis is the rather minor role that vegetation and soil appear to play in streamflow spatial variability at the scale of our inquiry. We use our perceptual model to build a distributed yet parsimonious hydrological model in Part 2 of this paper series. Although the perceptual model is specific for the Moselle, we hope that the perceptual model approach used here can be helpful to others working in large catchments in different hydroclimates, where other factors will dominate.