A statistical method based on the combination of singular spectrum analysis (SSA) with autoregressive statistical model is applied to reconstruct extended-range forecasts of precipitation from 1961 to 2005 in Guangdong. The results show that the precipitation in Guangdong reveals a 40-60 d circle and the correlation coefficient between reconstruction and observation is above 0.96; the forecast results of the other 4 a are good except the fifth pentad forecast in 2003. The highest correlation coefficient between extended-range forecast and observation of the first pentad is up to 0.86 and the fifth pentad is above 0.3. The average relative mean square errors are smaller than 30% for the 5 a, which means the prediction is of high credibility and stability by using SSA-AR and has reference value in extended-range forecast.