Evaluation of subseasonal forecast models for a strong heat wave over Southern South America

MARIANO S. ALVAREZ; MARISOL OSMAN

Qingdao

Conferencia; CLIVAR OSC 2016; 2016

CLIVAR

The prediction skill of subseasonal forecast models is evaluated for a strong and long-lasting heat wave occurred in December 2013 over Southern South America. Reforecasts from two models participating in the WCRP/WWRP Subseasonal to Seasonal project, the Bureau of Meteorology Poama and Beijing Climate Center model (CMA), were considered to evaluate their skill in forecasting temperature and circulation anomalies during that event. The Poama reforecast of 33 member ensemble size, initialized every five days, and CMA reforecast of 4 members ensemble size for each day were considered for the 1994-2013 period. CMA ensemble was constructed using the forecast of the same date of Poama plus the previous 4 days resulting in a 20 ensemble-members forecast. Weekly forecasts were computed with leadtimes from 2 days up to 22 days every five days. For each model, 2 meter temperature, 200-hPa and 850-hPa geopotential height and outgoing longwave radiation were used to describe the regional circulation. Daily ERA Interim reanalysis of 2 meter temperature and 200-hPa and 850-hPa geopotential height from 1994 to 2013 were used as observations while outgoing longwave radiation was taken from NOAA interpolated satellite estimates. Weekly anomalies were calculated for observations from 13th of December to 31st of December 2013. Anomalies for both observations and reforecast were calculated with respect to their own climatology.Results show that both models forecast lower temperature anomalies than those observed during week 1 and week 2 of the heat wave while Poama model overestimate temperature anomalies for week 3. Poama model forecasts the first two weeks of the heat wave at long and short leadtimes better than CMA although week 3 anomalies were better forecasted by CMA, especially for long leadtimes. Strong positive circulation anomalies over Southern South America at high and low levels during week 1 and 2 were well forecasted by Poama more than 15 days in advanced while CMA produced skillful forecast just two days prior to the beginning of the event. Week 3 negative circulation anomalies were of the opposite sing in long lead forecast of both Poama and CMA.In summary, for this case, models skill in forecasting surface temperature resulted moderate at long lead times. However, their performance for regional upper and low level circulation leads to the question whether the used of these forecasted variables in combination to other dynamic and statistic forecast tools could improve skill.