引用本文: | 吴新? 苗铮, 郝元? 董利? 基于beta回归的迎?号杨树树干密度混合效应模型[J]. 北京林业大学学报.doi:10.12171/j.1000-1522.20220450 |
Citation: | Wu Xinhua, Miao Zheng, Hao Yuanshuo, Dong Lihu. Mixed effect model of stem density ofPopulus nigra × P. simoniibased on beta regression[J].Journal of Beijing Forestry University.doi:10.12171/j.1000-1522.20220450 |
?nbsp; 1变量间的相关关系
ρs为边材密度,ρh为心材密度,ρb为树皮密度,ρw为树干密度,t为树龄,H为树高,Dt为胸径平均生长量+i>Dg为林分平均胸径,DBH为胸径,HD为林分平均高。蓝色代表负相关,红色代表正相关+i>P< 0.05。Notes:ρsis the sapwood density,ρhis the heartwood density,ρbis the bark wood density,ρwis the the wood density of stem,tis the tree age,His the tree height,Dtis the average growth of diameter at breast height,Dgis the average diameter at breast height of stand,DBHis the average diameter at breast height, andHDis the average height of stand. Blue for negative correlation,and red for positive correlation,P< 0.05.
Figure 1.Correlation between variables
?nbsp; 2林分因子和解析木因子基本信息统计?/p>
Table 2.Statistics of stand variables and sampling trees
因子 Factors | 最小 Min. | 最大 Max. | 平均 Mean | SD | |
林分因子 Stand factors | 林分年龄/a Stand age/year | 10 | 26 | 18 | 5 |
平均胸径 Average diameter at breast height/cm | 13.3 | 30.6 | 20.6 | 5.3 | |
平均树高 Average tree height/m | 11.9 | 26.2 | 20.3 | 4.1 | |
株数密度/(株·hm?)Tree density/(tree·ha?) | 183 | 933 | 572 | 267 | |
树木因子 Tree factors | 树木年龄/a Tree age/year | 10 | 26 | 18 | 5 |
胸径 Diameter at breast height/cm | 6.2 | 37.5 | 21.6 | 6.4 | |
树高 Tree height/m | 8.9 | 28.9 | 21.1 | 4.2 | |
边材密度 Sapwood density/(g·cm?) | 0.27 | 0.53 | 0.36 | 0.03 | |
心材密度 Heartwood density/(g·cm?) | 0.26 | 0.50 | 0.33 | 0.03 | |
树皮密度 Bark wood density/(g·cm?) | 0.20 | 0.55 | 0.37 | 0.07 | |
树干密度 Wood density of stem/(g·cm?) | 0.27 | 0.51 | 0.36 | 0.03 |
?nbsp; 3边材、心材、树皮和树干密度混合效应beta回归模型的拟合优度比辂/p>
Table 3.Fitting statistics test of mixed-effect beta regression model for sapwood, heartwood, bark and stem density
部位 Part | 模型编号 Model No. |
随机效应 Random effect | 评价指标 Evaluation indexes | ||||||||||
Hr | Hr2 | Dt | t | DBH | ?lnL | AIC | BIC | Ra2 | RL | P | |||
边材 Sapwood | 0 | ∑/td> | ∑/td> | ∑/td> | ? 988.27 | ? 978.27 | ? 950.60 | 0.31 | |||||
1 | ∑/td> | −△ | ∑/td> | ? 316.71 | ? 304.71 | ? 289.71 | 0.50 | ||||||
2 | −△ | ∑/td> | ∑/td> | ? 417.79 | ? 405.79 | ? 390.79 | 0.53 | ||||||
心材 Heartwood | 0 | ∑/td> | ∑/td> | ∑/td> | ? 012.41 | ? 002.41 | ? 976.12 | 0.19 | |||||
1 | ∑/td> | −△ | ∑/td> | ? 237.85 | ? 225.85 | ? 210.85 | 0.41 | ||||||
2 | ∑/td> | ∑/td> | −△ | ? 324.63 | ? 312.63 | ? 297.63 | 0.46 | ||||||
3 | −△ | ∑/td> | ∑/td> | ? 291.24 | ? 279.24 | ? 264.24 | 0.44 | ||||||
4 | −△ | ∑/td> | −△ | ? 383.05 | ? 367.05 | ? 347.05 | 0.52 | 58.42 | < 0.001 | ||||
树皮 Bark | 0 | ∑/td> | ∑/td> | ∑/td> | ? 887.39 | ? 877.39 | ? 850.11 | 0.29 | |||||
1 | −△ | ∑/td> | ∑/td> | ? 069.36 | ? 057.36 | ? 042.36 | 0.44 | ||||||
2 | ∑/td> | ∑/td> | −△ | ? 106.94 | ? 094.94 | ? 079.94 | 0.46 | ||||||
3 | ∑/td> | −△ | ∑/td> | ? 030.85 | ? 018.85 | ? 003.85 | 0.42 | ||||||
4 | −△ | −△ | ∑/td> | ? 163.77 | ? 147.77 | ? 127.77 | 0.52 | 56.83 | < 0.001 | ||||
树干 Stem | 0 | ∑/td> | ∑/td> | ∑/td> | ? 377.79 | ? 367.79 | ? 340.53 | 0.36 | |||||
1 | ∑/td> | −△ | ∑/td> | ? 566.35 | ? 554.35 | ? 539.35 | 0.50 | ||||||
2 | ∑/td> | ∑/td> | −△ | ? 736.63 | ? 724.63 | ? 709.63 | 0.56 | ||||||
3 | −△ | ∑/td> | ∑/td> | ? 651.70 | ? 639.70 | ? 624.70 | 0.53 | ||||||
4 | ∑/td> | −△ | −△ | ? 840.39 | ? 824.39 | ? 804.39 | 0.63 | ||||||
5 | −△ | ∑/td> | −△ | ? 822.19 | ? 806.19 | ? 786.19 | 0.62 | 103.76 | < 0.001 | ||||
注:△代表在该变量上添加随机效应,−代表模型固定效应的自变量、i>Hr代表相对高,Hr2代表相对高的平方+i>Dt代表胸径平均生长量,t代表树龄+i>DBH代表胸径,−2lnL代表负二倍的对数似然值,AIC代表赤池信息准则,BIC代表贝叶斯信息准则,Ra2代表调整确定系数+i>RL代表似然比。Notes: stands adding random effects to this variable, stands the independent variables of the model fixed effect,Hrstands the relative height,Hr2stands the square of relative height,Dtstands the average growth of diameter at breast height,tstands the tree age,DBHstands diameter at breast height, ?lnLstands the ? log likelihood value, AIC stands the akaike information criterion, BIC stands the bayesian information criterion,Ra2stands the stands the adjusted certainty coefficient,RLstands the likelihood ratio. |
?nbsp; 4边材、心材、树皮和树干密度最优混合效应模型的固定效应参数估计值、随机效应方差协方差结构
Table 4.Fixed effect parameter estimates, random effect variance covariance-structure for optimal mixed-effects model of sapwood, heartwood, bark and stem density
参数 Parameter | 边材 Sapwood | 心材 Heartwood | 树皮 Bark | 树干 Stem | |
固定效应参数估计 Fitted parameters | $ \;{\beta }_{0} $ | ?.543 9***(0.016 3) | ?.598 3***(0.028 9) | ?.800 0***(0.039 6) | ?.611 5***(0.025 9) |
$ \;{\beta }_{1} $ | 0.471 8***(0.032 1) | 0.709 0***(0.059 2) | 1.678 4***(0.099 3) | 0.241 6***(0.037 6) | |
$ \;{\beta }_{2} $ | ?.201 2***(0.034 2) | ?.082 1**(0.026 1) | ?.320 6***(0.111 3) | ?.003 6***(0.000 1) | |
$ \;{\beta }_{3} $ | ?.065 3***(0.012 8) | ?.288 7***(0.046 9) | ?.005 1*(0.001 9) | 0.094 4**(0.031 7) | |
随机效应方差协方差结 Covariance-structure | $ {\sigma }_{1}^{2} $ | 0.016 1 | 0.423 4 | 0.003 3 | 0.020 7 |
$ {\sigma }_{2}^{2} $ | 0.601 8 | 0.018 8 | 0.000 0 | ||
$ {\sigma }_{21} $ | ?.478 6 | ?.002 1 | ?.000 3 | ||
模型方差 Model variance | 0.010 4 | 0.010 0 | 0.044 3 | 0.008 8 | |
$ {\mathrm{注}:\beta }_{0}\mathrm{代}\mathrm{表} $模型固定效应通式(式?3))中的截距? \;{\beta }_{1} $? \;{\beta }_{2} $? \;{\beta }_{3} $代表模型固定效应通式中自变量前的系数? {\sigma }_{1}^{2}{\mathrm{和}\sigma }_{2}^{2} $代表模型随机效应的残差方差,$ {\sigma }_{21} $代表模型随机效应的残差协方差?**代表P< 0.001?*代表P< 0.01?代表P< 0.05。Notes: $ \;{\beta }_{0} $ stands the intercept in the general formula of the fixed effect of the model (equation (13)), $ \;{\beta }_{1} $, $ \;{\beta }_{2} $ and $ \;{\beta }_{3} $ stands the coefficient before the independent variable in the fixed-effect general equation of the model, $ {\sigma }_{1}^{2}\;{\mathrm{a}\mathrm{n}\mathrm{d}\;\sigma }_{2}^{2} $ stands residual variance of model random, $ {\sigma }_{21} $ stands residual covariance of model random effects effects. ***standsP< 0.001, **standsP< 0.01, *standsP< 0.05. |
?nbsp; 5基础模型与混合效应模型检验指标的比较
Table 5.Comparison of test indexes between base models and mixed effect models
部位 Part | 基础模型 Base model | 混合效应模型 Mixed effect model | |||
MAE/(g·cm?) | MAPE/% | MAE/(g·cm?) | MAPE/% | ||
边材 Sapwood | 0.022 1 | 6.069 1 | 0.018 8 | 5.155 1 | |
心材 Heartwood | 0.022 3 | 6.584 6 | 0.019 6 | 5.728 1 | |
树皮 Barkwood | 0.046 7 | 13.271 2 | 0.041 0 | 11.751 0 | |
树干 Stem | 0.021 5 | 5.953 3 | 0.017 0 | 4.692 0 | |
注:MAE为平均绝对误差,MAPE为平均绝对百分比误差。Notes:MAE is mean absolute error, and MAPE is mean absolute percentage error. |
?nbsp; 6混合效应模型抽样方案的MAPE检验指标统讠/p>
Table 6.MAPE test index statistics of sampling plans for mixed-effects model
方案−抽样数野br/> Plan-sampling size | 边材 Sapwood | 心材 Heartwood | 树皮 Bark | 树干 Stem |
?0 | 6.069 1 | 6.582 3 | 13.275 2 | 5.949 2 |
?1 | 6.056 2 | 6.599 8 | 13.345 3 | 5.964 5 |
?2 | 6.055 2 | 6.563 9 | 13.303 0 | 5.946 1 |
?3 | 6.050 6 | 6.547 5 | 13.288 3 | 5.937 7 |
?4 | 6.046 6 | 6.536 1 | 13.272 3 | 5.933 4 |
?5 | 6.045 6 | 6.534 1 | 13.272 2 | 5.930 8 |
?6 | 6.045 4 | 6.530 6 | 13.266 7 | 5.930 1 |
?1 | 5.968 7?. 0(/td> | 6.517 3?.3(/td> | 13.232 9?.0(/td> | 5.918 2?.0(/td> |
?2 | 5.966 0?, 1.0(/td> | 6.474 1?, 1.3(/td> | 13.232 4?, 2.0(/td> | 5.926 1?.0, 1.3(/td> |
?3 | 5.968 7?, 1.0, 1.3(/td> | 6.400 0?, 1.0, 1.3(/td> | 13.238 9?, 2.0, 1.3(/td> | 5.949 1?.0, 1.3, 2.0(/td> |
?4 | 6.040 4?, 1.0, 1.3, 2.0(/td> | 6.670 0?, 1.0, 1.3, 2.0(/td> | 13.249 7?, 1.0, 1.3, 2.0(/td> | 5.950 8?, 1.0, 1.3, 2.0(/td> |
注:表内括号中的内容代表取样圆盘高度(m)。Note: The content in parentheses in the table stands the height of the sampling disc (m). |