One of the most exciting prospects of using horizontal resolutions at kilometer-scale is the explicit numerical simulation of moist convection. However, convection is an intrinsically chaotic dynamical process with limited predictability. In this study we investigate how predictability of precipitation at meso-beta and smaller scales depends upon the type of precipitation and location with respect to topography as well as upon temporal and spatial scales.

To this end, ensemble experiments are performed using the Canadian non-hydrostatic high resolution numerical weather prediction model MC2 in a slightly modified MAP-SOP convection-resolving set-up. The applied ensemble methodology involves conducting six similar weather simulations with slightly modified initial conditions, but identical lateral boundary conditions. This set-up allows to isolate predictability aspects relating to meso-beta and meso-gamma scale variability. The modified initial conditions are generated by shifted initialization techniques.

Results are presented for MAP IOP3, MAP IOP15 and for a summer-day with strong thermal convection. The analysis suggests that the predictability heavily depends upon the weather type under consideration. For the spatial and temporal scales considered, the uncertainties in precipitation forecasts increase rapidly with decreasing scale and it is demonstrated how individual convective cells are rendered unpredictable by chaotic aspects of the moist dynamics.