APP下载

基于BP神经网络的酒泉市霜冻天气预测研究

2021-06-15王海燕鲁岳陈丽娜

河南科技 2021年5期

王海燕?鲁岳?陈丽娜

摘 要:本文利用2019年12月和2020年1月至11月气象观测资料、ECMWF再分析资料和“FY-4”陆表温度资料,基于BP神经网络对甘肃省酒泉市一次霜冻天气进行预报分析研究。研究表明:此次霜冻天气前期受到贝加尔湖西部冷槽影响使气温下降,后期乌拉尔山高压脊东移使天气转晴,冷平流及强辐射的共同影响是此次霜冻天气的主要原因;霜冻天气过程中,500 hPa最大北风带的建立、东移和南伸是一个极为重要的因素;基于BP原理,制作未来时刻的地面最低温度预报,确定此次霜冻天气过程的开始时间和影响范围,但霜冻天气预报对霜冻天气范围的计算有待提高。

关键词:BP神经网络;酒泉市;霜冻天气

中图分类号:P457.6文献标识码:A文章编号:1003-5168(2021)05-0144-03

Abstract: Based on the meteorological observation data in December 2019 and from January to November 2020, ECMWF reanalysis data and "FY-4" land surface temperature data, a frost weather forecast in Jiuquan City of Gansu Province was analyzed and analyzed based on BP neural network. The results show that: in the early stage of the frost weather, the cold trough in the west of Baikal Lake made the temperature drop, and in the later stage, the high pressure ridge of Ural Mountain moved eastward to make the weather clear. The combined effect of cold advection and strong radiation is the main cause of the frost weather; in the process of frost weather, 500 The establishment, eastward movement and southward extension of HPA maximum north wind zone is a very important factor; based on BP principle, the ground minimum temperature forecast in the future is made to determine the start time and influence range of the frost weather process, but the calculation of frost weather range in frost weather forecast needs to be improved.

Keywords: BP neural network;Jiuquan;frost

一直以來,陆地地表温度(Land Surface Temperature,LST)在地表与大气能量交换中扮演着重要角色[1-3],并且是地表过程分析和模拟的关键参数。它与植物的分布和生长、农作物、地表水资源蒸发循环、气候变迁、全球环境变化等有重要的关系[4]。目前,全球变暖引发极端天气出现的概率逐渐增大,在作物生长过程中,极端低温对其危害较大[5]。酒泉市是一个以农业为主的城市,地形复杂,在热量分布上存在显著性差异,霜冻现象比较常见。霜冻是一种对农业影响突出的灾害性天气,常常导致春季处在幼苗期的农作物、处于开花期的果树、处于灌浆期的大秋作物及晚熟品种遭到冻害,造成作物叶面损伤或枯萎,严重的可造成减产甚至绝产[6-8]。马尚谦提出甘肃省初霜冻日期、终霜冻日期、无霜冻日数分别遵循“北早南迟,西早东迟”“北迟南早,西迟东早”“北短南长,西短东长”的空间分布规律[9]。……

登录APP查看全文