基于Bayes判别法的马尾松毛虫一代、二代幼虫发生期的预报

Forecast for the occurrence period of Dendrolimus punctatus larvae based on Bayes discriminant method for the first and second generations

为了提高马尾松毛虫Dendrolimus punctatus(Walker)发生量预测预报结果的准确性,本文运用Bayes判别分析法建立安徽省潜山县1983年-2016年33年的马尾松毛虫一代和二代幼虫发生期的预报模型。一代幼虫发生期的判别函数方程为:f(1)=-15744.058-361.501x1+60.759x2+133.502x3+511.368x4;f(2)=-16854.938-375.596x1+70.405x2+132.608x3+529.690x4;f(3)=-17645.295-384.956x1+73.601x2+134.955x3+541.782x4;f(4)=-18179.639-382.408x1+71.342x2+135.234x3+549.655x4对1983年-2018年一代幼虫发生期预报结果历史符合率为97.06%,二代幼虫发生期的判别函数方程为:f(1)=-134898.483+559.235x5+113.112x6-250.033x7+1461.350x8;f(2)=-138908.622+573.572x5+118.340x6-252.691x7+1474.569x8;f(3)=-141430.680+577.358x5+125.727x6-254.610x7+1483.336x8;f(4)=-143185.175+578.968x5+129.628x6-256.102x7+1491.257x8对二代幼虫发生期的预报结果的历史符合率为100%。对2017年和2018年的验证回报,与实况结果一致。筛选出对预报量有密切关系的预报因子是本方法预报准确性的关键,该方法是一种简便准确性高的预报方法。

To improve the accuracy of forecasting the occurrence of Dendrolimus punctatus Walker,the Bayes discriminant analysis method was used to predict the occurrence period of the first and second generations of D.punctatus larvae over a period of 33 years from 1983 to 2016 in Qianshan county,Anhui province.The discriminant function equation of the occurrence period of the first-generation larvae was as followed:f(1)=-15744.058-361.501x1+60.759x2+133.502x3+511.368x4;f(2)=-16854.938-375.596x1+70.405x2+132.608x3+529.690x4;f(3)=-17645.295-384.956x1+73.601x2+134.955x3+541.782x4;f(4)=-18179.639-382.408x1+71.342x2+135.234x3+549.655x4.The historical coincidence rate of the forecast results from 1983 to 2018 was 97.06%.The discriminant function equation for the second-generation larvae was as followed:f(1)=-134898.483+559.235x5+113.112x6-250.033x7+1461.350x8;f(2)=-138908.622+573.572x5+118.340x6-252.691x7+1474.569x8;f(3)=-141430.680+577.358x5+125.727x6-254.610x7+1483.336x8;f(4)=-143185.175+578.968x5+129.628x6-256.102x7+1491.257x8.The historical coincidence rate of the forecast results for the second-generation larvae from 1983 to 2018 was 100%.The verification returns for 2017 and 2018 were consistent with the observed data.Screening out the forecasting factors closely related to the forecasting quantity was the key to the accuracy of forecast.This forecasting method is simple and accurate.