华南低能见度天气特征及客观预报研究

Study on low visibility weather features and objective forecast in South China

利用2000-2016年华南219个县级气象观测站的地面、高空气象观测资料以及对应站点的再分析资料,统计发生低能见度天气的天气形势和特征,归纳低能见度天气的预报指标。将与能见度以及能见度变化相关的气象要素输入神经网络进行训练,利用EC集合预报数据集获得能见度集合预报结果,通过对其离散度的统计分析以及经验公式最终获得具有泛用性、可靠性的神经网络模型的参数集。通过输入EC确定场数据,获得华南219县级站长时效精细化能见度预报结果,2017年上半年的能见度预报试验显示,模型预报结果的误差与TS评分均优于CUACE模式能见度预报。

The meteorological data during 2000 to 2016 from 219 observation meteorological stations of South China were collected for the research.The multi-timescales variation characteristics and the relations between visibility and meteorological elements were studied to summarize the weather conditions of low-visibility weathers.The selecting factors related to visibility and visibility change were input into ANN for training.The EC ensemble prediction data were used to obtain the forecast result and ANN model reliable.The long-term and meticulous visibility forecast of 219 stations in south China were calculated through the EC determined prediction data.The test result in the first half year of 2017 showed that the error and TS scores of the model prediction results were better than the CUACE mode visibility forecasting.