林业科学  2019, Vol. 55 Issue (3): 72-78   PDF    
DOI: 10.11707/j.1001-7488.20190308
0

文章信息

张雄清, 王翰琛, 鲁乐乐, 陈传松, 段爱国, 张建国.
Zhang Xiongqing, Wang Hanchen, Lu Lele, Chen Chuansong, Duan Aiguo, Zhang Jianguo.
杉木单木枯损率与初植密度、竞争和气候因子的关系
Tree Mortality in Relation to Planting Density, Competition and Climate Factors for Chinese Fir Plantation in Southern China
林业科学, 2019, 55(3): 72-78.
Scientia Silvae Sinicae, 2019, 55(3): 72-78.
DOI: 10.11707/j.1001-7488.20190308

文章历史

收稿日期:2017-02-14
修回日期:2017-05-26

作者相关文章

张雄清
王翰琛
鲁乐乐
陈传松
段爱国
张建国

杉木单木枯损率与初植密度、竞争和气候因子的关系
张雄清1,2, 王翰琛1, 鲁乐乐1, 陈传松3, 段爱国1, 张建国1,2     
1. 中国林业科学研究院林业研究所 国家林业和草原局林木培育重点实验室 北京 100091;
2. 南京林业大学南方现代林业协同创新中心 南京 210037;
3. 中国林业科学研究院亚热带林业实验中心 分宜 336600
摘要:【目的】分析杉木单木枯损率与初植密度、竞争和气候因子的关系,为杉木科学经营管理提供决策依据。【方法】以江西杉木密度试验林为研究对象,选取常用的logit、probit和cloglog 3种二分类变量数据结构模型构建杉木单木枯损率基础模型,并进行选择。以选择出的最优模型为基础,引入样地和样木的随机效应构建杉木单木枯损率混合效应模型。【结果】logit模型的AIC值最小(4 700.419),probit模型次之,cloglog模型最差。考虑样地和样木两水平随机效应的混合效应模型模拟精度最高,其AUC值为0.966 8。初植密度、林分优势高越大,杉木单木枯损率越高;相对直径d/Dg越大,杉木单木枯损率越低;气候越干旱,杉木单木枯损率越高;温度升高,杉木单木枯损率减小。【结论】考虑样地和样木两水平的logit模型能够较好分析杉木单木枯损率与初植密度、竞争、立地和气候因子的关系,并且随着气候干旱发生,杉木单木枯损率提高。
关键词:单木枯损率    杉木    连接模型    初植密度    气候变量    竞争    
Tree Mortality in Relation to Planting Density, Competition and Climate Factors for Chinese Fir Plantation in Southern China
Zhang Xiongqing1,2, Wang Hanchen1, Lu Lele1, Chen Chuansong3, Duan Aiguo1, Zhang Jianguo1,2     
1. Key Laboratory of Tree Breeding and Cultivation of the National Forestry and Grassland Administration Research Institute of Forestry, CAF Beijing 100091;
2. Co-Innovation Center for Sustainable Forestry in Southern China, Nanjing Forestry University Nanjing 210037;
3. Experimental Center of Subtropical Forestry, CAF Fenyi 336600
Abstract: 【Objective】It is important to analyze the relationship between tree mortality and planting density, competition, site index, and climate factors for managing Chinese fir(Cunninghamia lanceolata)well.【Method】Based on the Chinese fir data in Jiangxi Province, three widely used model for analyzing binary data which are logit, probit and cloglog link models were used to modelling tree mortality and compared with each other. Then we introduced the plot and tree random effects to the selected model in order.【Result】The result showed that the logit model was the best(AIC=4 700.419), followed by probit model, and cloglog model. Also accounting for the plot and tree random effects at the same time had the best performance, and its AUC was 0.966 8, which was close to 1. In addition, we also found:mortality rate increased with increasing initial planting density and site index; decreased with increasing d/Dg; decreased with increasing mean annual temperature and previous summer mean temperature; increased with increasing previous winter mean minimum temperature and annual heat-moisture index.【Conclusion】Logit model performed well on modeling tree mortality in relation to planting density, competition, site index, and climate factors. The research of relationship between mortality and annual heat-moisture index could be helpful for managing Chinese fir plantations under the climate change.
Key words: tree mortality    Chinese fir    link models    initial planting density    climate    competition    

树木枯损是指在环境干扰和遗传特性共同作用下发生的树木生命力逐渐减弱直到死亡的过程(Lee, 1971), 其受多种因素协同作用,且具有一定的物种特异性,是一个高度随机的过程(Franklin et al., 2002Yang et al., 2003)。根据树木死亡原因不同,枯损可分为自然枯损和非自然枯损2种(Vanclay, 1994)。其中,自然枯损主要是指树木生长发育过程中,由于树木自然成熟以及种内种间对林分内光照、水分、营养的竞争造成部分竞争力弱的树木逐渐枯死(Peet et al., 1987);非自然枯损主要是指由于外界随机因素(如冰冻雪灾、森林火灾或大风天气等)的干扰(Kneeshaw et al., 1998)造成树木逐渐枯死。当前,在枯损研究中涉及最多的是单木枯损(Holzwarth et al., 2013), 但是由于单木枯损影响因素的多样化和不确定性,使得学者们准确预测单木枯损相对较难,因此对于单木枯损的研究也主要局限在自然枯损方面。

影响单木枯损的因素可以归纳为内部因素和外部因素2大类。内部因素中种内种间竞争是导致枯损的主要原因(Monserud et al., 1999Franklin et al., 2002),初植密度不同,则竞争强度不同,影响单木枯损也不同。Lutz等(2006)研究发现,初植密度与竞争所导致的枯损相关性显著。气候是导致枯损的主要外部因素(Breshears et al., 2005Adams et al., 2009),温度和降雨量使得树木生长的气候环境发生改变,从而引发树木生长不适应性而发生枯损,特别是气候变化所导致的干旱天气(Mueller et al., 2005Zhang et al., 2014)。Allen等(2010)研究全球干旱天气与树木枯损的关系,认为干旱对全球树木生存是个巨大威胁。当然,树木枯损对气候变化的响应在不同树木直径结构、立地生产力和初植密度等林分也不同(Lutz et al., 2006Merlin et al., 2015Zhang et al., 2015)。

分析单木枯损率与诸因素的关系并准确估计枯损率,对维持种群动态变化、生态系统生物多样性和结构多样性等具有重要作用,由此出现了多种模型,如指数方程(Moser,1972)、Weibull方程(Somers et al., 1980)、Richards方程(Buford et al., 1985)、Gamma方程(Kobe et al., 1997)、对数正态分布方程(Preisler et al., 1997)和Hazards模型(Woodall et al., 2005)等。随着对单木枯损率模型研究的深入,学者们发现logit(logistic回归)模型分析单木枯损有较好的适用性(Vanclay,1995Zhang et al., 2011Boeck et al., 2014),且作为一个二分类变量数据结构(0或1)的单木枯损研究,probit或cloglog模型(complementary log-log)也是一种选择(Biggeri et al., 2001)。logit和probit模型具有一个共同点,即在0.5点对称,而cloglog模型却不对称,且取值1比0更快些。据此,Rose等(2006)Fortin等(2008)利用cloglog模型分析单木枯损率,结果也比较理想。当前,虽然这3种模型都做了一些研究,但究竟哪个模型更适用于分析单木枯损率还没有较好的探索比较。

杉木(Cunninghamia lanceolata)是我国亚热带地区特有的优良用材树种,也是我国南方主要的造林树种,第七次全国森林资源清查表明,杉木人工林面积为853.86万hm2,占全国造林面积的21.35%,在我国森林资源中占有重要地位(张雄清等,2014)。本研究以杉木人工林为研究对象,分别构建logit、probit和cloglog 3种单木枯损率模型并进行选择,分析初植密度、竞争、立地条件和气候等因素对杉木单木枯损率的影响,以期为杉木科学经营管理提供决策依据。

1 试验地概况及数据整理

试验地设在江西省分宜县大岗山年株林场场部后山,属罗霄山脉北端武功山支脉,地理位置114°30′—114°45′E,27°30′—27°50′N。林场场部后山海拔250 m,母岩为砂页岩,年均气温16.8 ℃,降雨量1 656 mm,年蒸发量1 503 mm,属南亚热带季风气候区。

试验林使用1年生苗木于1981年造林,采用随机区组试验设计,5种密度:2 m×3 m、2 m×1.5 m、2 m×1 m、1 m×1.5 m、1 m×1 m,每种密度3次重复,共15块样地,每块样地面积600 m2。在每块样地周围各栽植2行相同密度的杉木作为保护带。样地1989年前逐年调查,1989年后隔年调查。对样地内每株树编号,并进行每木检尺。本研究数据从1983到1999年,杉木林分的主要生长指标统计量见表 1

表 1 杉木林分及林木各变量统计值 Tab.1 Summary statistics of stand and tree variables of Chinese fir plantations
2 研究方法 2.1 变量选择

一般在选择变量时,不仅要考虑变量的T检验值,还要根据树木枯损可能潜在的影响因素进行分析(Adame et al., 2010)。Hamilton(1986)将引起树木枯损的内部因素分为3类:1)描述树木大小的变量,如胸径(d)、树高(h)等。对于多数树种来说,样地中较小的树木其枯损率较高,随着树木生长到大径阶其枯损率迅速减少(Adame et al., 2010)。Yao等(2001)认为,树木大小也可以用林龄(A)表示。2)描述竞争的一些指标,如林分断面积(B)、林分密度(N)、相对直径(d/Dg)、相对树高等。由于本研究树高没有全部测量,故相对树高未体现在模型中。3)立地条件,如林分优势木平均高(H)。研究发现,不同的立地条件水平,树木的枯损率也不同(Eid et al., 2001Yao et al., 2001)。此外,为了分析初植密度(initial planting density, IPD)对单木枯损率的影响,本研究中也将初植密度作为一个自变量;同时,由于初植密度数值较大,在进行模型分析前先对初植密度进行标准化处理。

气候变量采用Climate AP软件(Wang et al., 2014)获得, 主要有年平均温度(MAT,℃)、年降雨量(MAP,mm)、最冷月平均温度(MCMT,℃)、干燥指数(AHM)、零度以下天数(DD_0)、无霜冻天数(ND)、夏季平均最高温度(SMMT,℃)、冬季平均最低温度(WMMT,℃)和夏季平均温度(SMT,℃)。其中,干燥指数综合年平均温度和年降雨量2个指标,能够很好描述一年中的气候干燥情况:AHM=(MAT+10)/(MAP/1 000)(Zhang et al., 2014),AHM越大说明气候越干燥。为了消除各变量间的多重共线性,本研究采用逐步回归筛选自变量。

2.2 模型形式

对于1株给定的树,枯死与否可用0或1表示,这便形成了一个二分类变量数据结构。对于该结构,可由下式表示:

$ p\left({y = 1|x} \right) = F\left({x' \beta } \right)。$ (1)

式中:y为因变量,取值0或1;x为自变量;β为参数向量;F为累积分布函数。

累积分布函数不同,模型形式就不同,主要有logit(logistic)、probit和cloglog 3种模型(Wang et al., 2010)。结合本研究所用到的自变量,3种模型如下:

$ {\rm{logit}}:{\rm{ln}}\left({\frac{p}{{1 - p}}} \right) = f\left(x \right) + h({\rm{IPD}}) + g\left({气候} \right) + \varepsilon ; $ (2)
$ {\rm{probit}}:\;{\mathit{\Phi} ^{ - 1}}\left(p \right) = f\left(x \right) + h({\rm{IPD}}) + g\left({气候} \right) + \varepsilon ; $ (3)
$ {\rm{cloglog}}:{\rm{ln}}[ - {\rm{ln}}\left({1 - p} \right)] = f\left(x \right) + h({\rm{IPD}}) + g\left({气候} \right) + \varepsilon 。$ (4)

基于上述3种模型,本研究首先利用AIC值选择最优基础模型,模型AIC值最小,则该模型最优。然后以选择出的最优模型为基础,引入样地和样木的随机效应构建杉木枯损率混合效应模型,分析杉木单木枯损率与初植密度、竞争和气候因子的关系。在引入样地和样木的随机效应时,通过似然比检验(LRT)判断其必要性并通过AIC值选择模型。最后通过ROC曲线下方的面积(AUC)值判断单木枯损率估计精度。ROC曲线是根据一系列不同的二分类方式(分界值或决定阈),以真阳性率(灵敏度)为纵坐标、假阳性率(1-特异度)为横坐标绘制的曲线,AUC值越大,说明模型预测精度越高(Fielding et al., 1997)。AUC值传统判别方法为:0.9~1,优秀(A);0.8~0.9,较好(B);0.7~0.8,好(C);0.6~0.7,差(D);0.5~0.6,预测失败(F)。本研究计算在R软件中完成。

3 结果与分析

采用逐步回归筛选自变量后,得到logit、probit和cloglog 3种模型的参数估计和模型统计量见表 2。由表 2可知,各模型的参数估计值均在0.05水平上显著,logit模型的AIC值比probit模型小176.742,比cloglog模型小450.706, logit模型模拟杉木单木枯损率表现最优,probit模型次之,cloglog模型最差。因此,选择logit模型作为最优基础模型。

表 2 3种模型的参数估计值和模型统计量 Tab.2 Parameter estimates(standard deviation) and evaluation statistics of logit, probit and cloglog models

以logit模型为基础,先后引入样地和样木的随机效应构建杉木单木枯损率混合效应模型,模型的参数估计值见表 3。由表 3可知,考虑样地和样木两水平的模型AIC值最小,考虑样地单水平的模型AIC值次之,且都比不考虑随机效应的模型AIC值小;根据LRT检验,单水平和两水平的随机效应模型差异显著(P < 0.001)。由此可知,同时考虑样地和样木两水平的logit模型最优。

表 3 杉木枯损率混合效应模型的参数估计值 Tab.3 Parameter estimates and evaluation statistics of mixed models

通过模型参数估计结果发现,相对直径d/Dg的系数为负,表明竞争力越强,杉木单木枯损率越低;而初植密度(IPD)和林分优势高(lnH)相反,表明随着初植密度和林分优势高增加,杉木单木枯损率升高。对于气候因子,年平均温度(MAT)、夏季平均温度(SMT)与杉木单木枯损率负相关,而年干燥指数(AHM)和冬季平均最低温度(WMMT)则与杉木单木枯损率正相关(表 23)。

由ROC曲线(图 1)发现,考虑样地和样木两水平的模型AUC值为0.966 8,大于单水平的AUC值(0.955 4),且2种模型的AUC值都大于0.9,由此可知,logit混合效应模型用于预测杉木单木枯损率表现优秀。

图 1 杉木单木枯损率模型ROC曲线(左边为两水平模型,右边为单水平模型) Fig. 1 ROC of tree mortality model for Chinese fir(from left to right, they are two-level and one-level models, respectively)
4 讨论 4.1 logit、probit和cloglog模型的选择

logit、probit和cloglog是二分类变量数据结构常用的模型,广泛应用于单木枯损率研究中(Fortin et al., 2008Adame et al., 2010Zhang et al., 2011)。本研究利用这3种模型分析杉木单木枯损率与初植密度、竞争和气候因子的关系,结果发现,logit模型模拟效果最好,probit模型次之,cloglog模型最差。对于logit模型,其最大优点是参数的可解释性和简单操作性。Huettmann等(2003)研究表明,logit模型在推断和预测上优于probit和cloglog模型。Laaksonen(2006)采用logit、probit、log-log和cloglog模型分析居民消费倾向,结果发现logit模型并不总是最优,但始终比cloglog模型好。Raftery(1988)Kuson等(2012)也对这几个模型进行了比较分析,但结论不一。Vanclay(1995)建议采用logit模型分析热带森林单木枯损率。本研究以logit模型为最优基础模型,先后引入样地和样木的随机效应,结果表明,两水平的logit模型能够较好分析杉木单木枯损率与初植密度、竞争和气候因子的关系。混合效应模型可提高模拟精度已经在很多研究中得到了证实,而且两水平混合效应模型要优于单水平模型(Calama et al., 2005Yang et al., 2009符利勇等,2015),与本研究结论一致。

4.2 单木枯损率与竞争、立地和气候因子的关系

竞争是导致单木枯损的主要原因之一。本研究采用相对直径反映竞争,结果发现,相对直径越大,单木枯损率越低,与Laarmann等(2009)所得结论相一致。当然,林分断面积和林分株数密度也是反映竞争的重要指标,但在本研究分析过程中已通过逐步回归排除了这2个变量。对于初植密度与单木枯损率的关系,一般认为初植密度越大的林分,对水分、营养和光照的竞争越激烈,林分中的单木枯损率越高(Williams,1994Zhao et al., 2007);而且初植密度越大,树木的冠长和树冠率越小,由此也导致较高的单木枯损率(McClain et al., 1994Akerse et al., 2013)。对于立地与单木枯损率的关系,本研究发现,林分优势高越大,其单木枯损率越高,这在一些学者的研究中得到了证实(Eid et al., 2001Álvarez-González et al., 2004Zhang et al., 2015);然而,也有学者认为单木枯损率随着立地质量提高而减小(Woolons,1998Jutras et al., 2003)。因此,在分析单木枯损率与立地的关系时要谨慎。

气候是影响树木枯损的主要外界变量。本研究通过逐步回归,得到年平均温度(MAT)、夏季平均温度(SMT)、干燥指数(AHM)和冬季平均最低温度(WMMT)4个气候因子与单木枯损率在0.05水平上显著相关,其中,MAT、SMT与单木枯损率负相关,AHM、WMMT与单木枯损率正相关。Van Mantgem等(2009)Ruiz-Benito等(2013)研究发现,年平均温度升高,森林枯损率增加,与本研究结论不一致,可能是因为杉木属于喜光树种,温度小幅度上升,生长加快(吴中伦,1984)。干燥(干旱)是引起树木枯损的主要气候因子,气候越干燥,树木枯损率越高,本研究结论与大多数学者的研究结果相一致(Breshears et al., 2005Zhou et al., 2013Zhang et al., 2014)。

5 结论

本研究选取logit、probit和cloglog 3种二分类变量数据结构模型构建杉木单木枯损率基础模型,通过对3种模型的AIC值进行比较,得出logit模型是分析杉木单木枯损率与初植密度、竞争(相对直径)、立地(林分优势高)和气候关系的最优基础模型,以最优模型为基础,同时考虑样地和样木的随机效应,模型模拟精度最高,AUC值高达0.966 8。初植密度、林分优势高越大,杉木单木枯损率越高;相对直径越大,杉木单木枯损率越低。在气候因子中,年平均温度和夏季平均温度越高,杉木单木枯损率越低,而干燥指数和冬季平均最低温度越小,杉木单木枯损率越高。

参考文献(References)
符利勇, 唐守正, 张会儒, 等. 2015. 基于多水平非线性混合效应蒙古栎林单木断面积模型. 林业科学研究, 28(1): 23-31.
(Fu L Y, Tang S Z, Zhang H R, et al. 2015. Multilevel nonlinear mixed-effects basal area models for individual trees of Quercus mongolica. Forest Research, 28(1): 23-31. [in Chinese])
吴中伦. 1984. 杉木. 北京: 中国林业出版社.
(Wu Z L. 1984. Chinese fir. Beijing: China Forestry Publishing House. [in Chinese])
张雄清, 张建国, 段爱国. 2014. 基于贝叶斯法估计杉木人工林树高生长模型. 林业科学, 50(3): 69-75.
(Zhang X Q, Zhang J G, Duan A G. 2014. Tree-height growth model for Chinese fir plantation based on Bayesian method. Scientia Silvae Sinicae, 50(3): 69-75. [in Chinese])
Adame P, Del Río M, Cañellas I. 2010. Modeling individual-tree mortality in Pyrenean oak(Quercus pyrenaica, Willd.) stands. Annals of Forest Science, 67(8): 810-810. DOI:10.1051/forest/2010046
Adams H D, Guardiola-Claramonte M, Barron-Gafford G A, et al. 2009. Temperature sensitivity of drought-induced tree mortality portends increased regional die-off under global change-type drought. Proceedings of the National Academy of Sciences of the United States of America: 7063-7066.
Akers M K, Kane M, Zhao D, et al. 2013. Effects of planting density and cultural intensity on stand and crown attributes of mid-rotation loblolly pine plantations. Forest Ecology and Management, 310(1): 468-475.
Allen C D, Macalady A K, Chenchouni H, et al. 2010. A global overview of drought and heat-induced tree mortality reveals emerging climate change risks for forests. Forest Ecology and Management, 259(4): 660-684. DOI:10.1016/j.foreco.2009.09.001
Álvarez-González J G, Dorado F C, González A D R, et al. 2004. A two-step mortality model for even-aged stands of Pinus radiata D. Don in Galicia(Northwestern Spain). Annals of Forest Science, 61(5): 439-448. DOI:10.1051/forest:2004037
Biggeri L, Bini M, Grilli L. 2001. The transition from university to work:a multilevel approach to the analysis of the time to obtain the first job. Journal of the Royal Statistical Society:Series A, 164(2): 293-305. DOI:10.1111/rssa.2001.164.issue-2
Boeck A, Dieler J, Biber P, et al. 2014. Predicting tree mortality for European beech in southern Germany using spatially explicit competition indices. Forest Science, 60(4): 613-622. DOI:10.5849/forsci.12-133
Breshears D D, Cobb N S, Rich P M, et al. 2005. Regional vegetation die-off in response to global-change-type drought. Proceedings of the National Academy of Sciences of the United States of America: 15144-15148.
Buford M A, Hafley W L. 1985. Modeling the probability of individual tree mortality. Forest Science, 31(3): 331-341.
Calama R, Montero G. 2005. Multilevel linear mixed model for tree diameter increment in stone pine(Pinus pinea):a calibrating approach. Silva Fennica, 39(1): 37-54.
Eid T, Tuhus E. 2001. Models for individual tree mortality in Norway. Forest Ecology and Management, 154(1/2): 69-84.
Fielding A H, Bell J F. 1997. A review of methods for the assessment of prediction errors in conservation presence/absence models. Environmental Conservation, 24(1): 38-49.
Fortin M, Bédard S, Deblois J, et al. 2008. Predicting individual tree mortality in northern hardwood stands under uneven-aged management in southern Québec, Canada. Annals of Forest Science, 65(2): 205-210. DOI:10.1051/forest:2007088
Franklin J F, Spies T A, Van Pelt R, et al. 2002. Disturbances and structural development of natural forest ecosystems with silvicultural implications, using Douglas-fir forests as an example. Forest Ecology and Management, 155(1): 399-423.
Hamilton D A. 1986. A logistic model of mortality in thinned and unthinned mixed conifer stands of northern Idaho. Forest Science, 32(4): 989-1000.
Holzwarth F, Kahl A, Bauhus J, et al. 2013. Many ways to die-partitioning tree mortality dynamics in a near-natural mixed deciduous forest. Journal of Ecology, 101(1): 220-230. DOI:10.1111/jec.2012.101.issue-1
Huettmann F, Linke J. 2003. Assessment of different link functions for modeling binary data to derive sound inferences and predictions. Computational Science and Its Applications-ICCSA 2003: 43-48.
Jutras S, Hökkä H, Alenius V, et al. 2003. Modeling mortality of individual trees in drained peatland sites in Finland. Silva Fennica, 37(2): 235-251.
Kneeshaw D D, Bergeron Y. 1998. Canopy gap characteristics and tree replacement in the southeastern boreal forest. Ecology, 79(3): 783-794.
Kobe R K, Coates K D. 1997. Models of sapling mortality as a function of growth to characterize interspecific variation in shade tolerance of eight tree species of northwestern British Columbia. Canadian Journal of Forest Research, 27(2): 227-236. DOI:10.1139/x96-182
Kuson S, Sriboonchitta S, Calkins P. 2012. Household determinants of poverty in Savannakhet, Laos:Binary choice model approach. The Empirical Econometrics and Quantitative Economics Letters, 1(3): 33-52.
Laaksonen S. 2006. Does the choice of link function matter in response propensity modelling?. Model Assisted Statistics and Applications, 2(2): 95-100.
Laarmann D, Korjus H, Sims A, et al. 2009. Analysis of forest naturalness and tree mortality patterns in Estonia. Forest Ecology and Management, 258(supp.): S187-S195.
Lee Y. 1971. Predicting mortality for even-aged stands of lodgepde pine. Forestry Chronicle, 47(1): 29-32. DOI:10.5558/tfc47029-1
Lutz J A, Halpern C B. 2006. Tree mortality during early forest development:a long term study of rates, causes, and consequences. Ecological Monographs, 76(2): 257-275. DOI:10.1890/0012-9615(2006)076[0257:TMDEFD]2.0.CO;2
McClain K M, Morris D M, Hills S C, et al. 1994. The effects of initial spacing on growth and crown development for planted northern conifers:37-year results. Forestry Chronicle, 70(2): 174-182. DOI:10.5558/tfc70174-2
Merlin M, Perot T, Perret S, et al. 2015. Effects of stand composition and tree size on resistance and resilience to drought in sessile oak and Scots pine. Forest Ecology and Management, 339: 22-33. DOI:10.1016/j.foreco.2014.11.032
Monserud R A, Sterba H. 1999. Modeling individual tree mortality for Austrian forest species. Forest Ecology and Management, 113(2/3): 109-123.
Moser Jr J W. 1972. Dynamics of an uneven-aged forest stand. Forest Science, 18(3): 184-191.
Mueller R C, Scudder C M, Porter M E, et al. 2005. Differential tree mortality in response to severe drought:evidence for long-term vegetation shifts. Journal of Ecology, 93(6): 1085-1093. DOI:10.1111/j.1365-2745.2005.01042.x
Peet R K, Christensen N L. 1987. Competition and tree death. Bioscience, 37(8): 586-595. DOI:10.2307/1310669
Preisler H K, Slaughter G W. 1997. A stochastic model for tree survival in stands affected by annosum root disease. Forest Science, 43(1): 78-85.
Raftery A. 1988. Approximate bayes factors for generalized linear models. Technical reports No. 121, University of Washingtom.
Rose Jr C E, Hall D B, Shiver B D, et al. 2006. A multilevel approach to individual tree survival prediction. Forest Science, 52(1): 31-43.
Ruiz-Benito P, Lines E R, Gómez-Aparicio L, et al. 2013. Patterns and drivers of tree mortality in Iberian forests:climatic effects are modified by competition. PLoS ONE, 8(2): e56843. DOI:10.1371/journal.pone.0056843
Somers G L, Oderwald R C, Harms W R, et al. 1980. Predicting mortality with a Weibull function. Forest Science, 26(2): 291-300.
Van Mantgem P J, Stephenson N L, Byrne J C, et al. 2009. Widespread increase of tree mortality rates in the western United States. Science, 323(5913): 521-524. DOI:10.1126/science.1165000
Vanclay J K. 1995. Growth models for tropical forests:a synthesis of models and methods. Forest Science, 41(1): 7-42. DOI:10.1093/forestscience/41.1.7
Vanclay J K. 1994. Modeling forest growth and yield:applications to mixed tropical forests. School of Environmental Science and Management: 312.
Wang T, Wang G, Innes J, et al. 2014. Climate AP-an easy access to high-resolution climate data for Asia Pacific. Unpublished Manuscript, Faculty of Forestry, University of British Columbia, Vancouver, Canada.
Wang X, Dey D K. 2010. Generalized extreme value regression for binary response data:an application to B2B electronic payments system adoption. Annals of Applied Statistics, 4(4): 2000-2023. DOI:10.1214/10-AOAS354
Williams R A. 1994. Stand density management diagram for loblolly pine plantations in north Louisiana. Southern Journal of Applied Forestry, 18(1): 40-45.
Woodall W W, Grambsch P L, Thomas W. 2005. Applying survival analysis to a large-scale forest inventory for assessment of tree mortality in Minnesota. Ecological Modelling, 189(1/2): 199-208.
Woollons R C. 1998. Even-aged stand mortality estimation through a two-step regression process. Forest Ecology and Management, 105(1/3): 189-195.
Yang Y, Titus S J, Huang S. 2003. Modeling individual tree mortality for white spruce in Alberta. Ecological Modelling, 163(3): 209-222.
Yang Y, Huan S, Meng S X, et al. 2009. A multilevel individual tree basal area increment model for aspen in boreal mixedwood stands. Canadian Journal of Forest Research, 39(11): 2203-2214. DOI:10.1139/X09-123
Yao X, Titus S J, MacDonald S E. 2001. A generalized logistic model of individual tree mortality for aspen, white spruce, and lodgepole pine in Alberta mixed wood forests. Canadian Journal of Forest Research, 31(2): 283-291.
Zhang X, Lei Y, Cao Q V, et al. 2011. Improving tree survival prediction with forecast combination and disaggregation. Canadian Journal of Forest Research, 41(10): 1928-1935. DOI:10.1139/x11-109
Zhang X, Lei Y, Pang Y, et al. 2014. Tree mortality in response to climate change induced drought across Beijing, China. Climatic Change, 124(1): 179-190.
Zhang X, Lei Y, Liu X. 2015. Modeling stand mortality using Poisson mixture models with mixed-effects. iForest-Biogeosciences and Forestry, 8: 333-338. DOI:10.3832/ifor1022-008
Zhao D, Borders B E, Wang M, et al. 2007. Modeling mortality of second-rotation loblolly pine plantations in the piedmont/upper coastal plain and lower coastal plain of the southern United States. Forest Ecology and Management, 252(1/3): 132-143.
Zhou G, Peng C, Li Y, et al. 2013. A climate change-induced threat to the ecological resilience of a subtropical monsoon evergreen broad-leaved forest in Southern China. Global Change Biology, 19(4): 1197-1210. DOI:10.1111/gcb.12128