The availability of data collected during the Convective and Orographically-induced Precipitation Study (COPS) provides a unique opportunity to study the role of orography in determining the predictability of convection. Accurate prediction of the location of convective initiation requires high-resolution grids which can represent the fine-scale terrain details. In this study, we conduct high-resolution simulations using the UK Met Office Unified Model over the COPS region to quantify the predictability of severe convective storms encountered during COPS. An ensemble approach is taken using initial and boundary data from the Met Office Global and Regional Ensemble Prediction System (MOGREPS). Results are presented for two cases: one where there is weak synoptic forcing and convection was triggered by local convergence; and a second where convection was triggered by a squall line produced in the outflow boundary of a weak Mesoscale Convective System (MCS). For the case with weak synoptic forcing the ensemble spread from an initial conditions ensemble is small. This case shows sensitivity to parameters such as soil moisture and horizontal diffusion, demonstrating that it is also important to capture the uncertainty in the surface forcing and the parameterisation schemes. In contrast, for the synoptically-driven case MOGREPS produces an ensemble that better encompasses the observations.