Abstract: To investigate and access the short-term climate forecast techniques it is necessary to know how much of interannual variability of the monthly or seasonal mean quantities is potentially predictable. It is generally accepted that the total interannual variability can be partitioned into a potentially predictable component, which arises primarily from the persistence forcing by the lower boundary conditions of atmosphere and an unpredictable component induced by "weather noise". The ratio of the two components variance provides the measure of potential predictability. By means of daily precipitation data set the predictable climate signal and weather noise variance of seasonal precipitation in summer over the Guangxi are estimated in terms of low frequency white noise extension method and the analysis of variance under the assumptions of independence and dependence. Results show that there are potentially predictable climate signals over the region, the most strong, most weak climate signals are in the center and the east, the south of the region respectively. With the absolute error smaller than 0.68 standard deviation as the criterion of correct prediction, the upper limit of correctness would be 72%, 59%, 62% in the center and the east, the south, the north and the west respectively.