广西寒露风开始期短期气候预测方法研究

Short-term climate forecast method research on the start of cold dew wind in Guangxi

利用1951-2013年广西90个气象观测站气温资料、国家气候中心74项指数和美国National Oceanic and Atmospheric Administration(NOAA)的Climate Prediction Center(CPC)60项指数以及海温和陆地雪盖资料、美国国家冰雪研究中心(NSIDC)的两极海冰资料,使用相关分析方法得到广西寒露风开始期气候影响因子,利用逐步回归和神经网络方法进行寒露风开始期的预测。结果表明:寒露风开始期与前一年9-10月北极海冰面积、当年3月南极海冰面积、前一年6月欧亚雪盖、当年5月北美雪盖、北半球雪盖的相关显著。与前一年9月北半球极涡面积指数、前一年10月亚洲区极涡面积指数、前一年3月热带印度洋海温偶极子等指数相关显著。粒子群-神经网络方法预测误差低于逐步回归方法,预报能力有明显提高。

Based on the temperature data of 90 stations in Guangxi, the 74 indexes from National Climate Center（NCC）, the 60 indexes and the SST index data and the land snow data from CPC of NOAA during 1951-2013, the sea ice data from national snow and ice research center during 1979-2013, a new climatic prediction method of the onset of cold dew wind is studied by using the Fuzzy Neural Network method. The results show that： there are significantly correlations between the start of cold dew wind and the Arctic sea ice area from June to September in the year before, Antarctic sea ice in March in the same year, Eurasia snow cover in June in the year before, the North American and the Northern Hemisphere snow cover in May. The start of cold dew wind is also associated with some indexes, such as Arctic Oscillation（AO） index in September and October. The Particle Swarm Optimization（PSO） of Fuzzy Neural Network has a better qualitative capability of predicting the start of cold dew wind than the stepwise regression procedure.