天然红松林结实量预测模型的初步构建  

Preliminary Construction on Prediction Model of Nut Yield of Natural Korean Pine

作　　者：王君[1] 于海洋 徐宝福[3] 王元兴 李奎全 刘颖 周春艳 及利 杨雨春[1] Wang Jun;Yu Haiyang;Xu Baofu;Wang Yuanxing;Li Kuiquan;Liu Ying;Zhou Chunyan;Ji Li;Yang Yuchun(Jilin Academy of Forestry,Changchun 130033,China;Hongshi Forestry Bureau of Jilin Province,Huadian 132405,China;Huinan Forest Management Bureau of Jilin Province,Huinan 135102,China;Lushuihe Forestry Bureau of Jilin Province,Baishan 134506,China;Forestry Technology Institute of Jilin Province,Jiangyuan 134700,China)

机构地区：[1]吉林省林业科学研究院,吉林长春130033 [2]吉林省红石林业局,吉林桦甸132405 [3]吉林省辉南森林经营局,吉林辉南135102 [4]吉林省露水河林业局,吉林白山134506 [5]吉林省林业技师学院,吉林江源134700

出　　处：《北华大学学报（自然科学版）》2018年第5期588-594,共7页Journal of Beihua University（Natural Science）

基　　金：吉林省重点科技攻关项目(20170204003NY)

摘　　要：以露水河林业局天然红松林为对象,基于在16块固定样地中测定的677株红松单株结实量和林木因子的实测数据,应用非线性回归方法建立预测模型,并进行评价.结果表明:胸径和树高与结实量存在显著的正相关关系(P<0.01),分枝高和冠幅与结实量未达到显著相关,但与胸径和树高关系显著(P<0.01).以胸径与树高、胸径与分枝高和胸径与冠幅的不同因子组合建立结实量预测模型.通过对7个模型的R^2、均方根误差(RMSE)、赤池信息准则(AIC)、平均绝对偏差(MAE)、平均相对偏差绝对值(MRAE)等拟合优度和检验参数的比较分析,选择模型y=a×(D^2×BH)~b作为天然红松林结实量的最优模型,其预测精度为89.35%,消除异方差后的残差分布均匀;模型y=a×(D^2×H)~b和y=a×(D^2×CW)~b虽然R^2和P略高,但其平均偏差(ME)和MRAE与最优模型相差较大;以D^2作为因子建立的模型y=a×(D^2×H)~b、y=a×(D2×BH)b和y=a×(D^2×CW)~b皆表现出良好的检验效果.Nonlinear regression method was developed for predicting nut yields for natural Korean pine （ Pinus Koraiensis ） based on 16 fixed sample plots and determined the data of 677 Korean pine nut yield and forest factors in Lushuihe Forestry Bureau,then evaluated the prediction model.The result indicated that the diameter at breast height （DBH） and tree height （H） had a positive relation with nut yield significantly （ P 〈0.01）,while the branch height （BH） and crown width （CW） was not.But the BH had a significantly relationship with DBH and H （ P 〈0.01）.The prediction models of nut yield were established by different factors combinations,for example DBH and H,DBH and BH,DBH and CW.We compared and analyzed the goodness of fit and validation parameter such as R 2 ,root-mean-square error （RMSE）,akaike information criterion （AIC）,mean absolute error （MAE）,mean relative absolute error （MRAE） in every models,selected the model y=a×（D 2 ×BH） b as the optimal model of nut yield prediction which included the prediction accuracy of 89.35% and evenly distributed model residuals.However,the model y=a×（D 2 ×H） b and y=a×（D 2 ×CW） b had a larger R 2 and P ,but they were different from the optimal model about mean variation （ME） and average relative deviation absolute value （MRAE）.We found that the model y=a×（D 2 ×H） b ,y=a×（D 2 ×BH） b and y=a×（D 2 ×CW） b which made by the factor D 2 all indicated better test effect.

关 键 词：天然红松林 非线性回归模型 结实量 分枝高 

分 类 号：S722.1[农业科学—林木遗传育种]