1. Multi-model operational seasonal forecasts for SADC Willem A. Landman Asmerom Beraki Cobus Olivier Francois Engelbrecht

2. Conformal-Cubic Atmospheric Model (CCAM) Runs performed on a computer cluster at the University of Pretoria Runs performed on a computer cluster at the University of Pretoria Climatological ensemble runs - 12hr LAF (5 members) Climatological ensemble runs - 12hr LAF (5 members) Atmospheric initial conditions for climatological runs obtained from NCEP reanalysis data Atmospheric initial conditions for climatological runs obtained from NCEP reanalysis data Climatological simulations performed for the period: 1979-2003. Lower boundary forcing from AMIP SST and sea-ice Climatological simulations performed for the period: 1979-2003. Lower boundary forcing from AMIP SST and sea-ice

3. ECHAM4.5 at the SAWS All runs performed on NEC SX-8 All runs performed on NEC SX-8 Climatological (6 members) and operational ensemble runs - 24hr LAF Climatological (6 members) and operational ensemble runs - 24hr LAF Atmospheric initial conditions from ECMWF (1979 to 1996) analysis Atmospheric initial conditions from ECMWF (1979 to 1996) analysis Climatological dataset (1979-2003) constructed using AMIP physics; model constrained by lower boundary conditions generated from a high resolution AMIP2 dataset for SST and sea-ice Climatological dataset (1979-2003) constructed using AMIP physics; model constrained by lower boundary conditions generated from a high resolution AMIP2 dataset for SST and sea-ice Operational set-up: persisted and forecast SSTs obtained from a high resolution observed SST (optimum interpolation v-2) and IRI (mean) respectively (6 members each) Operational set-up: persisted and forecast SSTs obtained from a high resolution observed SST (optimum interpolation v-2) and IRI (mean) respectively (6 members each) 12-member ensemble operational runs on 18 th of each month for 6 consecutive months (i.e., 0-5 months lead- time) 12-member ensemble operational runs on 18 th of each month for 6 consecutive months (i.e., 0-5 months lead- time)

4. First objective multi-model forecast Old subjective consensus forecast

5. Combining algorithm: 1. CPT downscaling 2. Equal weights Multi-model ensemble Ensemble 1 (ECHAM4.5 at SAWS) 12 members Ensemble 2 (CCAM at UP) 5 members Ensemble 3 (CCM3.6 at IRI) 24 members Ensemble 4 (CFS at CPC) 40 members The current long-range forecast multi-model ensemble system of the South African Weather Service

6. New forecasting system UEA CRU data (0.5° resolution) UEA CRU data (0.5° resolution) –Precipitation –Minimum temperatures –Maximum temperatures MOS using 850 hPa geopotential height fields MOS using 850 hPa geopotential height fields –Domain: 10N-50S; 0-70E Production date: from July 2008

7. DJF rainfall simulation skill

8. DJF 1999/2000 precip & max temp PROBABILITY forecasts Precip Max T A typical example of the format of the forecasts

9. Rainfall forecast issued in December

10. DMC and VACS DMC DMC –SAWS to compile draft document on modernizing the SARCOF process –DMC has been receiving MM forecasts from SAWS since August 2008 MM work to be linked with VACS MM work to be linked with VACS –Workshop in 2009 (will introduce product)

11. ENSO forecast CCA (antecedent SST) CCA (antecedent SST) ECHAM4.5-MOM3 (from Dave DeWitt) ECHAM4.5-MOM3 (from Dave DeWitt) CFS (NCEP) CFS (NCEP)

12. Combining algorithm: 1. CPT downscaling 2. Equal weights Multi-model ensemble (& verification statistics) Ensemble 1 (ECHAM4.5 at SAWS) 12 members Ensemble 2 (CCAM at UP) 5 members Ensemble 3 (CCM3.6 at IRI) 24 members Ensemble 4 (CFS at CPC) 40 members The planned long-range forecast multi-model ensemble system of the South African Weather Service Ensemble 5+6 (+7) (GloSea4 at UKMO and CPTEC/COLA at INPE (ECMWF?))