We present an implementation of superparametrization (SP) of the ECMWF operational weather model OpenIFS by 3D high-resolution large-eddy simulations. Within a selected region, our software replaces the global model parametrizations of boundary layer turbulence, cloud physics and convection processes with tendencies derived from the vertical profiles of concurrently running instances of the Dutch Atmospheric Large Eddy Simulation (DALES). Vice versa, these DALES models are being forced towards the corresponding local vertical profiles of the OpenIFS. The coupling of the components has been implemented in a Python software layer using the OMUSE multi-scale physics framework. Our setup yields a "cloud-resolving" NWP model that has the potential to improve the representation of clouds and convection processes in OpenIFS. We analyze how the advection of liquid water through the SP region is underestimated and present a modifications in the SP scheme to resolve this. Finally we discuss scaling issues one faces when enlarging the SP region and possible future optimizations and load-balancing strategies to improve the performance of such a global SP.
