Are convection-permitting climate projections reliable for urban planning over Africa? A case study of Johannesburg

Cities are particularly vulnerable to surface water flooding. It is also well-known that they influence local rainfall themselves, which has important implications for climate change adaptation planning for cities. At km-scale resolution, convection-permitting climate models (CPCMs) better resolve cities and should better represent local urban temperature and rainfall modifications. However, using state-of-the-art pan-African CPCM simulations with the Met Office Unified Model (CP4), we show that for the city of Johannesburg, South Africa, this is not the case. A significant enhancement of rainfall occurs over the city compared with surrounding rural areas, which is not seen in available observations. We demonstrate this is associated with an overestimated urban heat island effect, which leads to additional triggering of rainfall. Urban signals in future rainfall change are small compared with changes in the wider surroundings, the latter of which we expect to be more reliable than in models with parameterized convection. This suggests that deficiencies in representation of urban processes are of secondary importance in terms of future percentage change in rainfall. We recommend urban planners apply relative changes in CP4 as an uplift to observations, where available, or treat absolute future rainfall as an upper estimate if used directly.