Abstract:
Global Climate Models (GCMs) and Regional Climate Models (RCMs) represent the atmospheric processes that are nonlinear by nature and are therefore sensitive to small perturbations. The RCMs are provided time dependent Lateral Boundary Conditions (LBCs) either from the GCM or the reanalyses and hence the RCMs are not expected to deviate much from the forcing fields as expected for a free non-linear system. If a GCM is used in a nested system, the nested solutions will be subject to the internal variability of both the GCM and the RCM. The study aims to investigate the variability caused by the internal variability of the GCM and the RCM. The study then looks into the contribution of the RCM’s internal variability to the total variability of the different nested system solutions. In this study four solutions obtained through perturbing the wind fields at initialisation for the ECHAM4.5 are used to force an RCM, the RegCM3, over South Africa. The solutions that are obtained are functions of the internal variability of the ECHAM4.5 as well as of the RegCM3. To determine the amount of the variability that is introduced by the RCM’s internal variability, four other RegCM3 simulations are made through initialising the RegCM3 on different days but using a single realisation from the GCM. The rainfall variability associated with the combined internal variability of both the models is high to an extent that ensemble members produce anomalies that have opposite signs in the same season. However, the sign of the ensemble average anomaly generally corresponds with the observed anomaly. The variability associated with the internal variability of the RCM is negligible when seasonal totals are analysed while with the daily rainfall totals the variability is larger. The variability in areas where small amounts of rainfall occur is smaller than that of the high rainfall regions. The number of events that fall into the three rainfall categories (i.e. below-normal, normal and above-normal) for the RegCM3 ensemble members are close to one another however the timing of the events is different. The results suggest that in operational forecasting making ensemble members associated with the internal variability of an RCM is not necessary because the information obtained from the ensemble members is almost similar.
