马尾松毛虫定量测报的判别分析模型

<正>马尾松毛虫的电子计算技术的定量测报,在我国还刚开始。本文应用二次判别分析,通过数字电子计算机对马尾松毛虫发生量和各种气象因子的相互关系进行分析。根据电算程序中的筛选标准自动筛选出影响马尾松毛虫发生的主要气象因子所组建的判别函数式,可作为测报数学模式。并据此测报出未来马尾松毛虫的发生量(共分为发生极轻,发生轻、发生重,发生极重四个量极)。判别分析电算定量测报方法应用于福建省连江县和浙江安吉县。

APPLICATION OF DISCRIMINANT ANALYSIS TO THE QUANTITATIVE FORECAST OF DENDROLIMUS PUNCTATUS WALKER

Dendrolimus punctatus Walker is one of the forest insect pest which does the greatest damage to the forest in China and is widely distributed over thirteen prov inces and one autonomous region of Guangxi, Guangdong, Fujian, Sichuan, Hunan, Jiangxi, Zhejiang, Henan, Hubei, Anhui, Jiangsu, Guizhou, Yunan and Taiwan with a yearly infected area up to 20-40 million mu, so it is of great significance to make good quantitative forecasts of Dendrolimus punctatus Walker, in order that comprehensive measures could be taken to eliminate it.The application of computerised techonology to the quantitative forecast of Dendrolimus punctatus Walker has now just begun in this country. In this paper, with digital computers and by the two-siage discrininant analysis, relationship between the incidence of Dendrolimus purciatus Walker and the weather factors is studied. According to the sieving standard of the program, primary meteorological factors influencing the incidence of Dendrolimus punclcius Walker are automatically sifted, which constitute discriminant functions that can be used as mathematical models for forecast and accordingly incidence of the future Masson-pirie caterpillar can be estimated (divided into 4 quantity classes; very low, low, high and extremely high). The method of discriminant analysis in the quantitative forecast by computerised means has been applied to the data from Lianjiang County, Fujian Province and Anji County, Zhejiang Province. The mathematical models used for forecasting the incidence of Dendrolimus punctatus Walker in Lianjiang County are as follows:The mathematical models used for forecasting the incidence on Dongfeng Farm in Anji County, Zhejiang Province are:Calculations show that these mathematical models reach a reliability of 84.6 -100% in the historical probability and good results can be obtained in forecasting by using them.