基于BCC-CSM季节气候预测模式系统历史回报数据和国家气象信息中心提供的中国地面降水月值数据,通过多方法对比并讨论了影响预测结果的因素,利用长短期记忆(Long Short-Term Memory,LSTM)网络预测2014年和2015年中国夏季降水.结果表明:LSTM网络的预测效果较逐步回归、BP神经网络及模式输出结果有一定优势.参数调优对于LSTM网络预测效果影响较大,重要参数有隐含层节点数、训练次数和学习率.选择合适的起报月份数据有助于提升季节预测的准确性,利用4月起报的数据预测夏季降水效果较好.海冰分量因子对降水季节预测有正贡献.在2014年、2015年夏季降水回报试验中,LSTM网络对降水整体形势有一定的预测能力,Ps评分分别为74分、71分,距平符号一致率分别为55.63％、55.25％,Ps评分的均值高于同期全国会商及业务模式.
1 地球系统科学需要人工智能 众所周知,地球系统包含了大气圈、水圈、陆地岩石圈、冰雪圈和生物圈等的复杂的相互作用和反馈,涵盖自然的和人类的全方位相互作用,直接涉及复杂的数学、地理、物理、化学和生物等多学科,间接则涉及社会和经济等众多领域.时间尺度可以从百万分之一秒到亿年,可以研究古代、历史时期、现代,以及未来的预测与预估;空间尺度宽泛,从原子到行星尺度,可以研究一个点,也可以研究整个宇宙,因此对认识地球系统带来许多困难.然而地球系统的变化涉及人类的生存,因此是全人类共同关心的问题.
Based on the definition of high temperature threshold and Heat-Wave Magnitude Index (HWMI),the temporal and spatial characteristics of temperature,heat wave frequency were analyzed by using the daily maximum temperature data from 716 observational stations in China during 1961-2014.Results show that the high temperature day in the area starting early (late) ends up late (early).Abrupt changes of high temperature days mainly happened between the end of twentieth century and the beginning of twenty-first century.Severe heat waves last from early July to early September and have great differences in each period of ten days.In Yunnan,most heat waves happen in May and seldom in other months.In addition to Huaihe River Basin,the tendency of heat wave frequency shows a wide range of positive value,especially in Guangxi,Guangdong,Yunnan and Hainan.HWMI index shows that index decreased progressively from 1960's to 1980's,while it increased progressively after 1990's.What's more,extreme and severe heat-wave has occurred more frequently since 1998,especially in the southern regionof Yangtze River.
