Quantifying internal variability in a regional climate model: A case study for Southern Africa

International audience This study aims at presenting various methodologies to separate the reproducible and irreproducible components of seasonal and intraseasonal climate variability simulated by a regional climate model over Southern Africa (south of 15°S), during an austral summer rainy season representative of the climatology. To that end, a 30-member ensemble simulation is performed using WRF laterally forced by the ERA40 reanalyses. Retained metrics include the analysis of weather regimes, signal-to-noise ratio, inter-member standard deviation and coefficient of variation. At the seasonal timescale, simulated rainfall amounts generally show a strong reproducibility, except in the subtropics and over the southern part of the Mozambique Channel. There, the number of rainy days is roughly similar in all members, while their average intensity varies extensively. At the intraseasonal timescale, the chronology of weather regimes, derived from the 500hPa geopotential height, is highly reproducible. Rainfall variability is much less reproducible, especially in the central parts of the domain and near its outflow boundaries. Analysis of a South African regional index nonetheless indicates that both wet and dry spells tend to be accurately simulated and occur in phase in most members, demonstrating that they are embedded in large-scale variability patterns. Internal variability is lastly related to the lateral forcings along the domain boundaries. An objective classification of inflow/outflow mass fluxes allows identification of the recurrent synoptic configurations that favor strong or weak regional reproducibility. The main uncertainties concern the basic morphological features of rain-bearing systems (i.e., their spatial extension, location and propagation speed). Consequences for tropical-temperate interactions are then discussed.