长白落叶松人工林优势高广义代数差分生长模型OA北大核心CSTPCD
Generalized Algebraic Differential Growth Model of Dominant Height for Larix olgensis Plantations
[目的]以长白落叶松人工林为研究对象,探究不同立地类型的立地质量差异,实现从立地分类到立地分级,精准评价长白落叶松人工林立地质量.[方法]基于吉林省一类调查和二类调查固定样地数据,采用广义代数差分法构建林分优势高生长差分模型.通过偏相关分析筛选对长白落叶松优势高生长具有显著影响的主导立地因子并形成立地类型(ST),应用K-Means算法对立地类型聚类形成立地类型组(STG),分别以ST、STG作为随机效应构建优势高生长混合效应模型,引入指数函数、幂函数、常数加幂函数消除异方差,采用一阶自回归、复合对称和自回归移动平均3种结构考虑自相关性,选出最优混合效应模型,根据立地类型组30年生的优势高进行立地分级.[结果]最优差分模型M1.1的调整决定系数(R2a)为0.865,相对均方根误差(rRMSE)为7.756%.对林分优势高具有显著影响(P<0.01)的立地因子(偏相关系数)为海拔(0.379)、土层厚度(-0.247)、腐殖质层厚度(0.190)、坡位(0.113)、坡度(0.067),前3项为主导立地因子,形成45个立地类型.45个立地类型聚类形成5个立地类型组,考虑立地类型组的混合效应模型M1.7最优,赤池信息准则(AIC)为4402.492、贝叶斯信息准则(BIC)为4 451.014,相较M1.1的AIC和BIC分别下降7.42%、7.04%.利用M1.7将5个立地类型组划分为5个立地等级,30年生的优势高依次为18.7、17.6、16.2、14.4和12.8 m.[结论]Richards、Korf、Hossfeld较其他基础模型更适宜推导差分模型;海拔、土层厚度、腐殖质层厚度是影响长白落叶松优势高生长的主导立地因子;获得立地类型与立地等级的对应表.
[Objective]Taking Larix olgensis plantations as the research object,this study aims to explore the difference of site quality between different site types,grade division of site from classification of site types is realized,and accurately evaluate the site quality of L.olgensis plantations.[Method]Based on the sample plot data by the national forest inventory and the forest management inventory in Jilin Province,a difference model for stand dominant height growth was constructed by generalized algebraic difference approach.The dominant site factors significant affecting dominant height growth of L.olgensis were screened by partial correlation analysis and formed the site types(ST).Site type groups(STG)is formed by clustering site types based on K-Means algorithm,and the ST and STG are taken as random effects to construct the dominant high growth mixed effect model respectively,adding exponential function,power function and constant plus power function to eliminate heteroscedasticity,autocorrelation of order 1,compound symmetry and autocorrelation-moving average,three structures were used to consider the autocorrelation,the optimal mixed effect model was selected and carried out the site rank according to the site type groups corresponding to the dominant height at 30 years.[Result]The optimal difference model M1.1 evaluation index adjusted coefficient of determination(R2a)=0.865,relative root mean square error(rRMSE)=7.756%.The site factors that had significant effect on stand dominant height(P<0.01)and their partial correlation coefficients were elevation 0.379,soil thickness-0.247,humus thickness 0.190,slope position 0.113,slope degree 0.067,respectively,the first three factors were identified as dominant site factors,and 45 site types were formed.5 site type groups were obtained from 45 site type clusters,the optimal mixed effect model M1.7 considering site type groups,AIC(Akaike information criterion)is 4 402.492 and BIC(Bayesian information criterion)is 4 451.014,decreased by 7.42%and 7.04%compared with AIC and BIC of M1.1 respectively.Divided into 5 site grades by using M1.7,the dominant height at 30 years are 18.7,17.6,16.2,14.4,12.8 m in turn.[Conclusion]Richards,Korf and Hossfeld are more suitable than other base models to derive difference models.Elevation,soil thickness and humus thickness are the dominant site factors affecting the growth of dominant height for L.olgensis.The corresponding table of site types and site grade is obtained.
罗光成;何潇;雷相东;吴璧芸;向玮
北京林业大学 林木资源高效生产全国重点实验室 北京 100083||中国林业科学研究院资源信息研究所国家林业和草原局森林经营与生长模拟重点实验室 北京 100091中国林业科学研究院资源信息研究所国家林业和草原局森林经营与生长模拟重点实验室 北京 100091中国林业科学研究院资源信息研究所国家林业和草原局森林经营与生长模拟重点实验室 北京 100091中国林业科学研究院资源信息研究所国家林业和草原局森林经营与生长模拟重点实验室 北京 100091北京林业大学 林木资源高效生产全国重点实验室 北京 100083
林学
广义代数差分法长白落叶松混合效应模型立地质量评价
generalized algebraic difference approachLarix olgensismixed effect modelsite quality evaluation
《林业科学》 2024 (12)
1-12,12
"森林立地质量评价与全周期多功能经营关键技术"(2022YFD2200500).
评论