摘要
本文主要对相同条件下线性回归法(LR)和人工神经网络法(ANN)对降雨、雨夹雪和降雪3种降水相态的预报效果进行了对比检验。选取降水发生时和发生前6 h的地面2m温度、露点温度作为预报因子,对降雨、雨夹雪和降雪进行预报。应用国家气象中心2001—2011年我国地面756站实况观测资料,其中应用2001—2010年资料对方法进行训练,2011年资料用来对比检验预报效果。结果显示,(1)两种方法对3种相态降水都有一定的预报能力,对降雪预报最好,其次是降雨和雨夹雪;(2)两种方法对北方的雨雪分界线预报比对南方的好;(3)无论是对全国还是长江中下游流域,在相同条件下,ANN法的预报效果大都优于LR法,当温度和露点温度预报准确时,ANN法对北方的雨雪分界线能进行较准确的预报。
The linear regression (LR) and artificial neural network (ANN) methods are compared with each other in forecasting precipitation types under the same conditions. The selected predictors are surface air temperature and dew point when and 6 hours before precipitation happens, and the types include rain, sleet and snow. The observation data from 756 weather stations of the National Meteorological Centre, CMA during 2001--2011 are used, of which the data of 2001--2010 are used to test the methods and the 2011 data are used to verify the forecasting effects. The results show that both of the LR and ANN meth- ods have prediction capacity for the three precipitation types of snow, rain and sleet. The predictability of snow is the best, then is rain, and the worst is sleet. Forecasts for the rain and snow separatrix forcasted by the two methods in the North of China are better than that in the South of China. The forecasting effect of ANN method is superior to that of LR method under the same conditions. When the temperature and dew point are forecasted correctly, the ANN method can be used to predict the rain and snow separatrix in the North of China exactly.
出处
《气象》
CSCD
北大核心
2013年第3期324-332,共9页
Meteorological Monthly
基金
公益性行业(气象)科研专项(GYHY201006010-2)
中国气象局气象关键技术集成与应用(1411140000004)
国家气象中心青年基金"模式预报温度偏差分析和误差订正"共同资助
关键词
人工神经网络
线性回归
降水相态
对比
artificial neural network (ANN), linear regression (LR), precipitation type, comparison