文献类型：期刊文献

中文题名：基于哑变量的日本落叶松生长模型研究

英文题名：Dummy variables model of increment of Larix kaempferi

作者：李忠国[1] 孙晓梅[1] 陈东升[1] 李子敬[1]

机构：[1]中国林业科学研究院林业研究所,国家林业局林木培育重点实验室,北京100091

年份：2011

卷号：39

期号：8

起止页码：69-74

中文期刊名：西北农林科技大学学报：自然科学版

收录：CSTPCD;;北大核心:【北大核心2008】；CSCD:【CSCD2011_2012】；

基金：林业公益性行业科研专项(201104027)

语种：中文

中文关键词：日本落叶松;树干解析;哑变量;生长模型

外文关键词：Larix kaempferi ; stem analysis ; dummy variable ; growth model

分类号：S758.2

摘要：【目的】研究北亚热带高山区和暖温带中山区日本落叶松的生长过程,并建立其生长模型,为精确掌握日本落叶松生长过程、科学经营日本落叶松人工林提供参考。【方法】应用北亚热带高山区和暖温带中山区96株日本落叶松解析木数据,根据区域特征引入哑变量的概念,综合运用Excel 2003、ForStat 2.1及SPSS 16.0等软件进行数据处理和生长模型的拟合,分别建立含有哑变量的日本落叶松胸径、树高和材积生长模型。【结果】2个区域日本落叶松的胸径、树高和材积生长情况没有明显差异。含有哑变量的Richards方程对日本落叶松胸径、树高生长拟合效果最好,R2分别达到了0.996 6和0.995 5,均方误差为0.163 2和0.207 7,平均绝对残差为0.349 1和0.436 7;材积生长模型的拟合结果以含有哑变量的二次函数最为理想,R2为0.997 979,均方误差为0.000 018,平均绝对残差为0.003 276。通过对模型的独立性检验,胸径、树高和材积生长模型预估精度均在90%以上。【结论】建立了日本落叶松胸径、树高和材积的哑变量生长模型,该模型可以用来描述北亚热带高山区和暖温带中山区日本落叶松的生长规律,预测其生长指标,解决了不同区域单独建模模型不相容的问题。
【Objective】 The growth model of the Larix kaempferi was constructed in northern sub-tropical alpine area and middle mountain area of warm-temperate zone to elucidate the growth process,so it can help us to manage the artificial forest of L.kaempferi more properly.【Method】 The concept of dummy variable was introduced according to the character of different areas while establishing the models of DBH,tree height and volume by stem analysis data of 96 trees.Excel 2003,ForStat 2.1 and SPSS 16.0 were used in processing data and curve fitting.【Result】 The results showed that the dummy variable Richards equation was the best model of the DBH and tree height,the determination coefficient R2 respectively achieved 0.996 6,0.995 5;the mean error MSE respectively was 0.163 2,0.207 7,the average absolute residual respectively was 0.349 1,0.436 7.The volume model’s fitting result by the dummy variable quadratic function was most ideal,with the determination coefficient R2 0.997 979,mean error MSE 0.000 018,the average absolute residuals 0.003 276.【Conclusion】 The models can be used actually for the accuracy is above 90% according to independent test.And the un-compatible problem of models built separately was resolved.