暴雨落区预报的映象方法

本文提出了一种暴雨落区预报的新方法。这种方法设想以最有利于产生暴雨区域的物理量数值为标准,全面考察其他地区相应物理量对产生暴雨的相对有利程度,用非线性映象的原理进行综合并数量化,最后将结果反射到天气图上,从而客观地作出暴雨落区预报。初步试验结果表明,该方法能客观地综合各物理量的主要分布特征,图面清晰,对暴雨落区的诊断与预报具有一定的能力。 在因子处理方面,我们用回归分析的方法,由计算机自动判别确定急流轴的位置;对映象的计算过程,我们根据最优化指标K的函数特性,用梯度法代替过去一般采用的简单松驰迭代算法,克服了由于样本个数增加,简单松驰迭代算法不收敛的缺点,取梯度模长的平方小于万分之一作为迭代收敛的标准,改进后一般迭代次数不超过50次,达到了业务使用的要求。

THE IMAGE PREDICTION FOR TORRENTIALRAIN AREA

A new method for prediction of torrential rain area is presented.It stand- ardizes those valuses of atmo-physical parameters in the area where it is the most favourable to form torrential rain; Inspects by homologues elsewhere,in all round,the relatively favourable level of producing torrential rain.The level is synthesized and numericalized by the principle of nonlinear mapping,and reflected on to a synoptic chart.Thus a predictional chart of torrential rain area can be located objectively and clearly.In dealing with factors,we used the computer to distinguish and determine the location of jet axis automaticaly by means of the regression analysis.For calculating process,we replace the method of simple relaxational iteration with that of gradient method according to the functional characteristics of the optimum index K,therefore overcome the default that simple relaxational iteration doesn't converge with the number of the sample increased.