Recent research has highlighted the crucial role for ecological communities and the significant prevalence of temporary streams: over 50% of the world's rivers are temporary. However, the main parameters and processes involved in the stream intermittency remain poorly understood. The literature identifies several controlling factors, including meteorology, geology and land cover, but a major challenge remains: understanding the role of the subsurface, i.e. the effect of the aquifer's hydraulic properties. In line with the new definition of an intermittent river, “a non-perennial river or stream with a considerable connection to the groundwater table, having variable cycles of wetting and flow cessation, and with flow that is sustained longer than a single storm event”, we use a 3D process-based groundwater flow model at the catchment-scale to simulate stream flow and the spatio-temporal extension/contraction of the hydrographic network. Focusing on two crystalline catchments in Brittany (northwest France), with a similar temperate climate, we explore the combined effect of hydraulic conductivity K (transmission capacity) and porosity θ (storage capacity). We propose a calibration approach with performance criteria based solely on surface information: 1) stream flow measured at the catchment outlet and 2) mapping of the observed hydrographic network (for the low-water period: perennial streams and for the high-water period: perennial + intermittent streams). The calibration results show that the methodology leads to an optimal aquifer model for the 2 catchments (Canut: K = 4.5 x 10-5 m/s, θ = 0.1% and Nancon: K = 1.5 x 10-5 m/s, θ = 2.2%). For both catchments, the comparison of simulated versus observed stream flow, focusing on low flows, achieves an excellent NSElog>0.86. At the same time, the model reproduces very well the hydrographic network observed during both high and low water periods (Figure 1), resulting in a very good success criterion. Finally, the models are validated using data from discrete visual monitoring of stream hydrological conditions (national ONDE network). Based on these modelling results, we then discuss hydrogeological controls such as the role of aquifer storage capacity on headwater stream intermittency. We believe that the use of stream intermittency mapping, benefiting from innovations in crowdsourcing and remote sensing, will address the need for hydrological models and predictions for ungauged basins.