1. Interannual Variability of the Rainy Season in the Tropics of South America Brant Liebmann NOAA-CIRES Climate Diagnostics Center Boulder, Colorado, USA José A. Marengo Centro de Previsão de Tempo e Estudos Climáticos Cachoeira Paulista, Brasil Anji Seth Suzana Camargo International Research Institute for Climate Prediction The Earth Institute at Columbia University Palisades, New York

2. Themes: a)The variability of the initiation (and of the end) of the rainy season is important for determining calendar season rainfall totals (when the initiation or end occurs during the season of interest). b) When there is a relationship between seasonal rainfall anomalies and sea surface temperature (SST), it is almost always through the relationship between SST and the variation of the starting or ending date.

3. Rainfall: 1976-1999 We wish to thank the following agencies: Agência Nacional de Energia Elétrica (Brasil) UADASC - AASANA - Santa Cruz (Bolivia) Ministerio del Ambiente y los Recursos Naturales (Venezuela) National Climatic Data Center (USA) FUNCEME, SIMEPAR, DAEE (Brasil) C.T.M. Salto Grande (Uruguay - Argentina) Administración Provincial de Agua (Argentina) Servicio Meteorológico Nacional (Argentina, Paraguay, Uruguay) Meteorologische Dienst Suriname METEO-France (French Guiana) UTE (Uruguay)

5. September-November climatology December-February climatology

6. March-May climatology June-August climatology

7. Niño 3.4 = SST in the region 5N - 5S 170  W -120  W Linear Correlation (simultaneous)

8. A(day) = Anomalous accumulation R(n) = daily rainfall R = annual average daily precipitation The rainy season is considered to be when the slope of the curve is positive ( R(n) > R ).

9. Average date of initiation: 30 December East Central Amazon

10. composite about initiation and end day relative to initiation or end rainfall (mm/day)

11. 1982-83 (6 March onset)

12. Month of initiation of rainy season (Climatology) october november december january february march april

15. Month of end of rainy season (from climatology) november october september august july june march april may

16. Correlations: Initiation versus DJF total: -0.86 Rate versus DJF total: 0.77 Rate versus initiation date: -0.56 …..in the east central Amazon Basin, the average date of onset is 30 December Standard Deviation: 28 days Therefore, usually the rainy season begins during DJF. Therefore, the variability of the day of initiation Is important in determining the DJF total. The rate of precipitation is defined as the total between the initiation and the end of February divided by the number of days.

17. DJF total rainfall versus SST Start versus SSTRate (start to end of February) versus SST

18. AMJ rain versus SST Rate (1 april to end of rainy season) versus SSTEnd date versus SST Average ending date = 13 june 2.5N-2.5S 60W-55W

19. 2S-4N 56W-49W Average ending date = 20 June May-July rain versus SST Rate (1 May to end) versus SSTEnding date versus SST

20. Summary: a)Variability of the start (and end) of the rainy season is important in determining the quantity of rain during the calendar season (when the start or end occurs during the season of interest). b) When the variability of total rainfall during the rainy season is related to SST, this relationship usually is due to the relationship between SST and the variation of the start (or end) of the rainy season.

21. General Circulation Model European Centre / Hamburg Model (ECHAM-3) (Used operationally at IRI until January 2002) 1 member (of 8) Run 50 years with observed SST (we use 1976-1997) Precipitation fields only Horizontal resolution: T42 Vertical resolution: 19 Levels Tiedke mass flux convection scheme Modified SiB land surface scheme

22. Same, except ECHAM-3 December-February Climatology 1976/77 – 1996/97

23. March-May ClimatologySame, except ECHAM-3

24. June-August Climatology Same, except ECHAM-3

25. September-November ClimatologySame, except ECHAM-3

26. oct nov dec jan feb mar apr GCM uses 360 day year

27. Note different scales

28. ~ during wet season: OBS GCM % dry days 7.0 8.4 rainy day average 10.1 9.0

29. Dec-Feb Obs. rain vs. Nino 3.4 SST Same, except ECHAM-3

30. Dec-Feb rain vs. nino 3.4 SST Regression for 1 std SST anomaly observed ECHAM-3

31. Note different scales

35. Grid point at 56W, 1.4S Observations GCM Average onset= 30 December 13 December Std of onset= 28 days 48 days Using ensembles may reduce standard deviation of onset date

38. We argue that the influence of SST on seasonal total precipitation is through its influence on the timing of the rainy season. Conclusions The same relationship may hold in a GCM, but if timing is biased or spread is erroneous, it will appear that the GCM does not properly simulate calendar season interannual variability. Perhaps bias and spread will be improved through utilization of ensembles. Statistical techniques (e.g., adjusting for bias) may improve interannual simulations.