Level and spatial pattern of overstory retention impose tradeoffs for regenerating and retained trees

Experimental design and treatment implementation The experiment is a randomized complete-block design (Aubry et al. 1999, Aubry and Halpern 2020). Here, we consider four of the treatment combinations representing one of two levels of overstory retention (15% vs. 40% of initial basal area) distributed in one of two spatial patterns (trees uniformly dispersed, D, or in intact circular patches or aggregates, Ap). We do not consider the unharvested controls. Harvest units are 13 ha in area and square or slightly rectangular (edge lengths of 320–400 m). In the dispersed treatments (15D and 40D), a portion of the dominant and co-dominant trees was retained in a uniformly dispersed pattern using, as the retention target, the cumulative basal area of the corresponding aggregated treatment (15A or 40A). The aggregated treatments contained either two or five 1-ha (56 m radius) patches. In 15A, patches were spaced ~115 m apart on the diagonal of the harvest unit; in 40A, the patches were spaced ~30 m apart. All merchantable stems (>18 cm at diameter breast height, dbh) were felled in the adjacent harvested area (15Ah or 40Ah) except those converted to snags (described below). Felling and yarding occurred over a period of 3-7 months in 1997 or 1998. Methods of yarding varied: helicopters were used on steep slopes (DP, B, and LW) and tracked shovel-loaders or rubber-tired skidders, on gentler terrain (WF and PH). Tree limbs were left attached to the bole to reduce slash accumulation. However, woody residues were deemed excessive at WF and were partially reduced by piling and burning (Halpern and McKenzie 2001). Treatment of non-merchantable stems (<18 cm dbh) varied among sites. Stems were retained at DP, B, and LW. In contrast, they were cut at PH and cut if damaged at WF, reducing the density of advance regeneration. To meet requirements of the Northwest Forest Plan (USDA and USDI 1994, Tuchmann et al. 1996), snags felled during harvest were replaced, in part, by artificially created snags. A total of 6.5 live trees/ha—typically large, decadent or broken-topped Pseudotsuga menziesii—were retained in the harvested portion of each unit, then were topped (B and PH) or girdled (WF, DP, and LW). To meet federal stocking requirements (312 stems/ha), 1- and 2-year-old bare-root seedlings (a mix of conifer species) were planted in the harvested portion of each unit in the spring or early summer after harvest. Planting densities were kept low (416–591 seedlings/ha) to minimize interactions with natural regeneration. We report on the growth and survival of planted trees in an earlier paper (Urgenson et al. 2013b), but not here. Timing and methods of sampling Pre-harvest measurements (1994-1996), were used to characterize initial forest structure and to establish the basal-area targets for dispersed-retention units. The first post-harvest measurement (year 1) was made in 1998 (B and PH) or 1999 (WF, DP, and LW); subsequent measurements were made in 2003 (year 5 or 6), 2009 (year 11 or 12), and 2016 (year 18 or 19). Prior to harvest, we established a systematic grid of 63 or 64 points (40-m spacing) in each experimental unit. Grid points serve as the centers of nested plots used to sample regeneration (>0.1 m tall and <5 cm dbh, including advance regeneration and post-harvest establishment) and overstory trees (≥5 cm dbh). Regeneration plots were distributed at alternate (n = 32) grid points in dispersed treatments and at a subset of grid points in the harvested areas of aggregated treatments (n = 22 in 15Ah; n =12 in 40Ah). In each regeneration plot we tallied all conifer and hardwood species in four 1 × 6 m belt transects oriented along perpendicular radii, 4 to 10 m from the plot center. Overstory plots (0.04 ha, 11.28 m radius) were distributed at alternate grid points in dispersed treatments (n = 32), at all grid points in the patches of aggregated treatments (n = 10 in 15Ap; n = 24-25 in 40Ap), and at the same grid points used to sample regeneration in the harvested areas of aggregated treatments (n = 22 in 15Ah; n =12 in 40Ah). In the first post-harvest year, all retained trees were identified to species, tagged, and measured for dbh (0.1 cm resolution). At each re-measurement we recorded dbh and ‘status’ (live or dead). For dead trees, we also recorded ‘position’ (i.e., standing intact, standing but broken, or down), allowing us to distinguish between wind and other causes of mortality (Urgenson et al. 2013a). At each re-measurement we tagged and recorded dbh of all ingrowth stems (i.e., trees attaining 5 cm dbh since the previous measurement). Data aggregation and processing Seral status and size class of trees. — We classified all conifers as early seral (ES) or late seral (LS) based on their shade tolerance or ability to regenerate under a closed canopy (Minore 1979, Burns and Honkala 1990). Hardwoods (uncommon at most sites) were not classified by seral status, but were included in any analyses that combined the responses of seral groups. In addition to seral status, retained trees (conifers and hardwoods) were grouped by size class: ‘small’ (<25 cm dbh; typically subordinate or subcanopy stems) and ‘large’ (≥25 cm dbh; typically dominants or co-dominants ranging to 171 cm dbh). This allowed us to compare growth and survival of trees of similar size and canopy position. Response variables. — For each plot within a harvested area (D or Ah), we computed final regeneration and ingrowth density by seral group, and for trees in total. For each plot with surviving ingrowth, we also computed the mean individual growth rate (annualized basal area increment; cm2/yr) of each seral group and for trees in total. For each plot with retained trees (D or Ap), we computed the mean individual growth rate of surviving trees of each size class (small and large) and seral group. For each harvest unit, individual growth rates were summed and expressed on area basis (m2/ha/yr) to quantify (1) ingrowth productivity in the harvested area (D or Ah), (2) retained-tree productivity in areas with retention (D or Ap), and (3) ingrowth and retained-tree productivity for the harvest unit as a whole (thus accounting for the differences in harvested and patch area between 15A and 40A). Finally, for each harvest unit we computed survival of retained trees of each size class and seral group. Ingrowth survival was consistently high (92%) and was not analyzed further. We computed two measures of variability among plots within harvest units: the coefficient of variation (CV) in density of regeneration and ingrowth (by seral group and total), and the ‘multivariate dispersion’ of plots (Anderson et al. 2006). Multivariate dispersion was computed as the mean Euclidean distance of plots to the harvest-unit centroid based on a principal coordinates analysis (PCoA) of a Bray-Curtis dissimilarity matrix of early- and late-seral regeneration and ingrowth density. Density values were natural-log transformed prior to generating the matrix. To facilitate inclusion of plots lacking regeneration and ingrowth, a pseudo-species with a minimum density was added to each sample. Multivariate dispersion was computed using the betadisper function in the vegan package in R (Oksanen et al. 2018). Post-harvest covariates. — We also computed a set of structural metrics to capture initial post-harvest (year 1) variation unaccounted for by the nominal retention treatments. Post-harvest advance regeneration density—which varied widely among sites, harvest units, and plots—was used in models of final regeneration and ingrowth density. Post-harvest CVs and multivariate dispersion, measures of spatial variability in advance regeneration density and composition within harvest units, were used in models of final variability. Density and basal area of early- and late-seral species, which varied widely among replicates of the same retention treatment, were used as treatment-scales proxies of seed availability in models of regeneration density. Finally, total density and basal area, plot-scale proxies of local tree influence, were used in models of retained-tree growth and survival. Literature cited Anderson, M. J., K. E. Ellingsen, and B. H. McArdle. 2006. Multivariate dispersion as a measure of beta diversity. Ecology Letters 9:683– 693. Aubry, K. B., M. P. Amaranthus, C. B. Halpern, J. D. White, B. L. Woodard, C. E. Peterson, C. A. Lagoudakis, and A. J. Horton. 1999. Evaluating the effects of varying levels and patterns of green-tree retention: Experimental design of the DEMO Study. Northwest Science 73 (Special Issue):12–26. Aubry, K. B., and C. B. Halpern. 2020. The Demonstration of Ecosystem Management Options (DEMO) Study, a long term-experiment in variable-retention harvests: rationale, experimental and sampling designs, treatment implementation, response variables, and data accessibility. General Technical Report PNW-GTR-978. USDA Forest Service, Portland, Oregon, USA. Burns, R. M., and B. H. Honkala, technical coordinators. 1990. Silvics of North America: 1. Conifers. Agriculture Handbook 654. USDA Forest Service, Washington, D.C., USA. Halpern, C. B., and D. McKenzie. 2001. Disturbance and post-harvest ground conditions in a structural retention experiment. Forest Ecology and Management 154:215–225. Minore, D. 1979. Comparative autecological characteristics of northwestern tree species - A literature review. USDA Forest Service General Technical Report PNW-GTR-087. Oksanen, J., F. G. Blanchet, M. Friendly, R. Kindt, P. Legendre, D. McGlinn, P. R. Minchin, R. B. O’Hara, G. L. Simpson, P. Solymos, M. H. H. Stevens, E. Szoecs, and H. Wagner. 2018. Vegan: Community ecology package. R package version 2.5-2. https://cran.r‑project.org/web/packages/vegan/vegan.pdf Tuchmann, E. T., K. P. Connaughton, L. E. Freedman, and C. B. Moriwaki. 1996. The Northwest Forest Plan: A Report to the President and Congress. USDA Forest Service, Portland, Oregon, USA. Urgenson, L. S., C. B. Halpern, and P. D. Anderson. 2013a. Level and pattern of overstory retention influence rates and forms of tree mortality in mature, coniferous forests of the Pacific Northwest, USA. Forest Ecology and Management 308:116–127. Urgenson, L. S., C. B. Halpern, and P. D. Anderson. 2013b. Twelve-year responses of planted and naturally regenerating conifers to variable-retention harvest in the Pacific Northwest, USA. Canadian Journal of Forest Research 43:46–55. USDA, and USDI. 1994. Record of Decision for Amendments to U.S. Forest Service and Bureau of Land Management Planning Documents Within the Range of the Northern Spotted Owl. USDA Forest Service, Portland, Oregon, USA.

实验设计与处理实施
本实验采用随机完全区组设计（randomized complete-block design）（Aubry等，1999；Aubry与Halpern，2020）。本研究选取4种处理组合，分别对应2种林冠保留（overstory retention）水平：初始断面积（basal area）的15%与40%，并设置2种空间分布模式：树木均匀分散（记为D），或完整圆形斑块/聚块（记为Ap）。本研究未纳入未采伐对照样地。采伐单元（harvest unit）面积为13 ha，形状为正方形或近似矩形（边长320–400 m）。

在分散保留处理（15D与40D）中，通过保留优势木与亚优势木，使保留木的累积断面积与对应聚块保留处理（15A或40A）的目标断面积一致，且分布呈均匀分散模式。聚块保留处理包含2个或5个1 ha（半径56 m）的聚块。在15A处理中，聚块沿采伐单元对角线间距约115 m；在40A处理中，聚块间距约30 m。

除被改造为枯立木（snags）的个体外，采伐区域（15Ah或40Ah）内所有商品材级木（胸径>18 cm，dbh，diameter at breast height）均被砍伐。采伐与集材（yarding）作业于1997年或1998年的3–7个月内完成。集材方式因地制宜：陡坡区域使用直升机（对应DP、B、LW样地），平缓地形则使用履带式铲运机或橡胶轮胎集材机（对应WF、PH样地）。树枝保留于树干上以减少采伐剩余物堆积，但WF样地的木质剩余物被认为过量，因此通过堆烧进行了部分清理（Halpern与McKenzie，2001）。

非商品材级木（胸径<18 cm，dbh）的处理方式因样地而异：DP、B、LW样地保留该类木；PH样地对其全部砍伐，WF样地仅砍伐受损个体，以此降低天然更新幼树（advance regeneration）的密度。为满足《西北森林计划》（Northwest Forest Plan，USDA与USDI，1994；Tuchmann等，1996）的要求，采伐过程中被伐倒的枯立木部分通过人工创建枯立木进行补充。每个采伐单元的采伐区域内共保留6.5株活立木/ha，通常为大型、衰退或顶梢折断的北美黄松（Pseudotsuga menziesii）。随后对这些保留木进行截顶（对应B与PH样地）或环剥（girdle，对应WF、DP与LW样地）。

为满足联邦立木株数达标要求（312株/ha），采伐后的春季或初夏，每个单元的采伐区域内种植1年生和2年生裸根实生苗（bare-root seedlings，针叶树混合种）。种植密度较低（416–591株实生苗/ha），以减少与天然更新植株的竞争。我们已在先前的论文中报道了种植木的生长与存活情况（Urgenson等，2013b），本文不再赘述。

采样时间与方法
采伐前调查（1994–1996年）用于表征初始森林结构，并确定分散保留处理的断面积目标。首次采伐后调查（第1年）于1998年（对应B与PH样地）或1999年（对应WF、DP与LW样地）开展；后续调查分别于2003年（第5或6年）、2009年（第11或12年）及2016年（第18或19年）进行。

采伐前，每个实验单元内设置63或64个系统样点（systematic grid，间距40 m）。样点作为嵌套样地（nested plots）的中心，用于调查更新植株（高度>0.1 m且胸径<5 cm，dbh，包括天然更新幼树和采伐后萌发生长的植株）与林冠木（胸径≥5 cm，dbh）。

更新样地设置于分散保留处理的交替样点（n=32），以及聚块保留处理采伐区域的部分样点（15Ah样地n=22；40Ah样地n=12）。每个更新样地内沿垂直半径方向设置4条1×6 m的带状样带（belt transects，距样点中心4–10 m），统计所有针叶树和阔叶树物种。

林冠样地面积为0.04 ha（半径11.28 m），设置于分散保留处理的交替样点（n=32）、聚块保留处理聚块内的所有样点（15Ap样地n=10；40Ap样地n=24–25），以及聚块保留处理采伐区域内与更新样地相同的样点（15Ah样地n=22；40Ah样地n=12）。

采伐后第1年，所有保留木均被鉴定物种、标记并测量胸径（精度0.1 cm）。每次重测时记录胸径与“状态”（活立木或枯立木）。对于枯立木，同时记录“位置”（即完整站立、站立但折断或倒伏），以此区分风倒与其他死亡原因（Urgenson等，2013a）。每次重测时，对所有新长成木（ingrowth stems，即上次测量后胸径达到5 cm的树木）进行标记并测量胸径。

数据聚合与处理
树木的演替地位与径级分类
我们根据耐荫性（shade tolerance）或在封闭林冠（closed canopy）下的更新能力，将所有针叶树划分为早期演替种（ES）或晚期演替种（LS）（Minore，1979；Burns与Honkala，1990）。阔叶树（多数样地中较为少见）未按演替地位（seral status）分类，但在整合演替组响应的分析中予以纳入。除演替地位外，保留木（针叶树与阔叶树）按径级（size class）分组：“小径级”（胸径<25 cm，通常为被压木或亚冠层木）与“大径级”（胸径≥25 cm，通常为优势木或亚优势木，最大胸径可达171 cm）。此举可用于比较相同大小和冠层位置树木的生长与存活情况。

响应变量
对于采伐区域（D或Ah）内的每个样地，我们计算了各演替组及总体的最终更新密度与新长成木密度。对于存在存活新长成木的样地，我们还计算了各演替组及总体的单株平均生长速率（年化断面积增长量，cm²/yr）。对于存在保留木的样地（D或Ap），我们计算了各径级（小径级与大径级）及演替组的存活木单株平均生长速率。对于每个采伐单元，将单株生长速率求和并以面积为基准（m²/ha/yr）进行表达，以此量化：(1) 采伐区域（D或Ah）的新长成木生产力；(2) 保留区域（D或Ap）的保留木生产力；(3) 整个采伐单元的新长成木与保留木生产力（因此考虑了15A与40A处理中采伐区域与聚块面积的差异）。最后，对于每个采伐单元，我们计算了各径级及演替组的保留木存活率。新长成木的存活率始终较高（92%），未进行进一步分析。

我们计算了采伐单元内样地间的2种变异性指标：更新密度与新长成木密度（按演替组及总体）的变异系数（CV，coefficient of variation），以及样地的“多变量离散度”（multivariate dispersion）（Anderson等，2006）。多变量离散度通过基于布雷-柯蒂斯相异矩阵（Bray-Curtis dissimilarity matrix）的主坐标分析（PCoA，principal coordinates analysis）计算样点到采伐单元质心的平均欧氏距离得到，该相异矩阵基于早期演替与晚期演替的更新密度及新长成木密度。在生成矩阵前，密度值已进行自然对数转换（natural-log transformed）。为纳入无更新与新长成木的样地，每个样本中均添加了具有最小密度的伪物种（pseudo-species）。多变量离散度通过R语言vegan包（vegan package）中的betadisper函数计算（Oksanen等，2018）。

采伐后协变量（covariates）
我们还计算了一系列结构指标，以捕捉名义保留处理未解释的采伐后第1年的变异。采伐后天然更新幼树密度（在样地、采伐单元间差异显著）被用于最终更新密度与新长成木密度的模型。采伐后变异系数与多变量离散度，即采伐单元内天然更新幼树密度与组成的空间变异性指标，被用于最终变异性的模型。早期与晚期演替物种的密度与断面积（同一保留处理的重复样地间差异显著）被用作更新密度模型中种子可用性的样地尺度替代指标。最后，总密度与断面积，即局域树木影响的样地尺度替代指标，被用于保留木生长与存活的模型。

参考文献
[1] Anderson, M. J., Ellingsen, K. E., McArdle, B. H. 2006. 多变量离散度作为β多样性的测度指标. Ecology Letters 9:683–693.
[2] Aubry, K. B., Amaranthus, M. P., Halpern, C. B., 等. 1999. 评估不同水平与模式的绿木保留：DEMO研究的实验设计. Northwest Science 73 (特刊):12–26.
[3] Aubry, K. B., Halpern, C. B. 2020. 生态管理选项示范（DEMO）研究：一项长期可变保留采伐实验的原理、实验与采样设计、处理实施、响应变量及数据可访问性. General Technical Report PNW-GTR-978. 美国农业部林业局，俄勒冈州波特兰.
[4] Burns, R. M., Honkala, B. H.（技术协调）. 1990. 北美林木学：1. 针叶树. Agriculture Handbook 654. 美国农业部林业局，华盛顿特区.
[5] Halpern, C. B., McKenzie, D. 2001. 结构保留实验中的干扰与采伐后地表条件. Forest Ecology and Management 154:215–225.
[6] Minore, D. 1979. 西北部树木物种的比较生态特性——文献综述. USDA Forest Service General Technical Report PNW-GTR-087.
[7] Oksanen, J., Blanchet, F. G., Friendly, M., 等. 2018. Vegan：群落生态学包. R包版本2.5-2. https://cran.r-project.org/web/packages/vegan/vegan.pdf
[8] Tuchmann, E. T., Connaughton, K. P., Freedman, L. E., 等. 1996. 西北森林计划：致总统与国会的报告. 美国农业部林业局，俄勒冈州波特兰.
[9] Urgenson, L. S., Halpern, C. B., Anderson, P. D. 2013a. 林冠保留的水平与模式对美国太平洋西北成熟针叶林树木死亡率的速率与形式的影响. Forest Ecology and Management 308:116–127.
[10] Urgenson, L. S., Halpern, C. B., Anderson, P. D. 2013b. 美国太平洋西北可变保留采伐下种植与天然更新针叶树的12年响应. Canadian Journal of Forest Research 43:46–55.
[11] USDA, USDI. 1994. 北斑猫头鹰分布范围内美国林务局与土地管理局规划文件修正案的决定记录. 美国农业部林业局，俄勒冈州波特兰.