Aboveground Wood Production Is Sustained in the First Growing Season after Phloem-Disrupting Disturbance

: Carbon (C) cycling processes are particularly dynamic following disturbance, with initial responses often indicative of longer-term change. In northern Michigan, USA, we initiated the Forest Resilience Threshold Experiment (FoRTE) to identify the processes that sustain or lead to the decline of C cycling rates across multiple levels (0, 45, 65 and 85% targeted gross leaf area index loss) of disturbance severity and, in response, to separate disturbance types preferentially targeting large or small diameter trees. Simulating the e ﬀ ects of boring insects, we stem girdled > 3600 trees below diameter at breast height (DBH), immediately and permanently disrupting the phloem. Weekly DBH measurements of girdled and otherwise healthy trees ( n > 700) revealed small but signiﬁcant increases in daily aboveground wood net primary production (ANPP w ) in the 65 and 85% disturbance severity treatments that emerged six weeks after girdling. However, we observed minimal change in end-of-season leaf area index and no signiﬁcant di ﬀ erences in annual ANPP w among disturbance severities or between disturbance types, suggesting continued C ﬁxation by girdled trees sustained stand-scale wood production in the ﬁrst growing season after disturbance. We hypothesized higher disturbance severities would favor the growth of early successional species but observed no signiﬁcant di ﬀ erence between early and middle to late successional species’ contributions to ANPP w across the disturbance severity gradient. We conclude that ANPP w stability immediately following phloem disruption is dependent on the continued, but inevitably temporary, growth of phloem-disrupted trees. Our ﬁndings provide insight into the tree-to-ecosystem mechanisms supporting stand-scale wood production stability in the ﬁrst growing season following a phloem-disrupting disturbance.


Introduction
The spatial footprint of disturbance from wood boring insect pests is immense, but the resulting effects of phloem disruption on carbon (C) storage in wood remains uncertain [1]. Each year in the United States alone, an estimated 2.4 Mha of forestland is invaded by phloem-disrupting insects [2], giving rise to large gradients of stand-scale disturbance severity [1,[3][4][5]. While C cycling responses to stand-replacing disturbance are well-studied, the mechanisms underlying whole ecosystem C cycling responses to less severe phloem-disrupting disturbances are poorly understood [2,[5][6][7][8][9]. Notably, phloem disruption may result in a more gradual reduction in tree growth relative to disturbances from fire, extreme Table 1. Predisturbance tree species composition and stem densities by canopy class, and total aboveground woody biomass for each of four treatment replicates ( Figure 1). Canopy classes are defined by diameter at breast height (DBH). The four most abundant canopy and seedling/sapling species, and the five most abundant subcanopy species, are provided. Species abbreviations are: Populus grandidentata (POGR), Quercus rubra (QURU), Acer rubrum (ACRU), Acer saccharum (ACSA), Fagus grandifolia (FAGR), Pinus strobus (PIST), Pinus resinosa (PIRE), Acer pensylvanicum (ACPE), Amelanchier (AMEL), and Viburnum acerifolium (VBAC). Standard errors for stem densities and biomass are presented parenthetically.
Forests 2020, 11 [30]. Treatment replicates serving as experimental blocks are distributed across four distinct forest ecosystems that are common to the region, and which are delineated above using the color scheme of Table 1. Lines distinguish between forest ecosystems varying by site productivity, soils, landform and species composition (A) [36]. Each replicate contains whole plots that were randomly assigned disturbance severities of 0, 45, 65 or 85% gross leaf area index loss and then split by top-down or bottom-up disturbance type (B).

Leaf Area Index
We estimated postdisturbance upper canopy leaf area index (LAI) from hemispherical images taken with a SONY Alpha 6000 24.3-megapixel DSLR camera with a 180-degree hemispherical lens. Images were taken 1 m above the forest floor at five nonoverlapping locations within each subplot beginning in May 2019 and continuing weekly through August 2019 [38] (FoRTE Project Data, https://fortexperiment.github.io/fortedata/). Images were processed for LAI using WinsCanopy (Regent Instruments, Quebec, Qu).

Aboveground Wood Net Primary Production
We derived daily and annual ANPP w for 2019 from repeated measurements of DBH (or, in the case of seedlings, basal diameter) [39]. During summer 2018, we censused upper canopy trees within each subplot, identifying individuals to the species level. In the same field season, we installed dendrometer bands at 1.3 m height on~25% of all canopy trees (n = 666 total) using a stratified random selection process to ensure our banded subsample spanned the species abundances and DBH distributions of each subplot population. Dendrometer bands were custom made using 1 2 inch steel tape, 2 inch stainless steel springs and circumference ruler stickers. Before installing dendrometer bands, a thin, outer layer of bark was mechanically shaved to achieve a smooth, tight fit around the stem. The DBH of banded trees was recorded during a one-year period on 17 separate one to two-day measurement campaigns on the following dates: November 2018; April 2019; weekly from May through August 2019 and November 2019.
Daily relative growth rates (RGRs, cm cm −1 day −1 ) of banded trees were estimated as the diameter growth increment of a time interval divided by the diameter at the start of the time interval. When species and plot-specific linear relationships between DBH and RGR were significant (p < 0.05) in a time interval, we modeled the RGR of the unsampled tree populations using regression equations generated from observations. When relationships between DBH and RGR were not significant, we estimated RGRs of the unsampled population from week, subplot and species-specific mean RGRs. From measured or RGR-inferred DBH, we estimated aboveground wood biomass of upper canopy stems using species and site or region-specific allometries [34,37]. With the goal of understanding how tree-level responses to girdling scale to influence the stand-level ANPP w , we separately compared the RGRs of girdled and healthy tree species using unpaired t-tests.
For subcanopy stems, we estimated density via species-level censuses conducted in one quarter of each subplot (0.025 ha). On six occasions from May 2019 to August 2019, we selected two individuals within or, if absent, closest to the center of each of four, 2-m 2 vegetation sampling areas per subplot. Stems were identified to the species level and their DBH measured using digital calipers. A quarter of the stems sampled for subcanopy DBH were contained within vegetation sampling areas, and all others were sampled outside of designated vegetation sampling areas but within treatment subplots. Mean subplot and species-specific DBH increments were applied to censused stems to infer their growth. Species and site or region-specific allometries [37] were used to calculate subplot aboveground wood biomass.
For seedling and saplings, we censused stems in one randomly selected quarter of each of the 2-m 2 vegetation sampling areas twice: once in June and again in August 2019. We recorded the species, basal diameter and current height, along with the height of each individual at the end of the prior growing season, from branch scars. Following the convention for measuring stems below DBH [40][41][42], the wood volume of each stem was calculated assuming conical geometry and species-specific wood densities were applied to convert volume to aboveground wood biomass [43].
From aboveground wood biomass increments, we estimated the mean daily ANPP w (kg C ha −1 day −1 ) of upper canopy trees and annual ANPP w (kg C ha −1 year −1 ) from the cumulative year-long sum of upper canopy, subcanopy, and seedling/sapling aboveground wood production. Daily ANPP w is upper canopy aboveground wood biomass increment expressed on a daily timestep, scaled to the hectare and multiplied by a site-specific C fraction of 0.48 to convert biomass to C mass [35]. Annual (upper canopy + subcanopy + seedling/sapling) ANPP w is the November 2018 to November 2019 increment of aboveground wood biomass from all stem size classes scaled to the hectare and converted to C mass. The uncertainty of each ANPP w estimate was calculated as the standard error of mean (daily or annual) ANPP w among replicates [35].
In advance of analyzing our results, we examined the sensitivity of our ANPP w estimates to potential stem swelling above girdled tissues [44], i.e., whether aboveground wood production was inflated because of disproportionately high radial growth at the location of DBH measurements, 30 cm above the removed bark. In early July 2019, on a subset of healthy and girdled trees, we installed a second set of tightly fitted upper dendrometer bands 50 cm above the already installed DBH-positioned bands (n = ≥3 per species). Beginning two weeks after installation, upper and DBH stem diameter increments were recorded concurrently for four weeks from July to August 2019 and again in November 2019. To determine whether disproportionately high radial growth occurred at DBH in girdled trees, we compared the taper of stem diameter increments from upper and lower (DBH positioned) bands in girdled and healthy trees using unpaired t-tests for each species. While moderately significant, Acer rubrum was the only species displaying signs of greater taper between the upper and lower growth increments in girdled trees (0.0064 mm day −1 ) relative to healthy trees (0.0021 mm day −1 ; p = 0.06; data not shown). To test the sensitivity of ANPP w to a potentially inflated diameter increment in A. rubrum, we compared unadjusted and adjusted (i.e., applied a taper adjustment to A. rubrum) estimates of aboveground wood production using unpaired t-tests, finding no significant differences between the two approaches (Table A4). Accordingly, our numeric presentation of results and statistical analysis used unadjusted ANPP w estimates.

Statistical Analysis
We used a time-series split-plot analysis of variance (ANOVA) model to analyze differences in daily canopy ANPP w among disturbance severity and disturbance type treatments. Disturbance severity (fully randomized within replicates) was the whole-plot and disturbance type (randomized within each whole-plot) was the split-plot (Table A1). For disturbance treatment comparisons of annual ANNP w , two similar split-split plot models, without time as a factor, were used to analyze differences among disturbance severities, disturbance types, girdled/healthy trees (Table A2) and successional cohort (Table A3) with either girdled/healthy or cohort considered a second (fixed) parameter in the model, respectively. We compared girdled and healthy tree RGR (pooled across disturbance severity and type treatments) and assessed the contribution of each to total annual ANPP w within disturbance severity and type treatments. Because our girdling treatments and subsequent high frequency (i.e., weekly) stem diameter measurements were concentrated in the upper canopy size class, we limited our analysis of daily ANPP w and annual ANPP w by girdled/healthy trees to the upper canopy. We included all canopy strata in the analysis of annual ANPP w by successional cohort to provide the most comprehensive assessment of successional cohort contributions to annual ANPP w . A priori pairwise comparisons of ANPP w across time, disturbance severity, disturbance type, successional cohort, and between girdled and healthy trees, were performed using least squares differences (LSD) (α = 0.05). Unpaired t-tests were used for pairwise comparisons of girdled and healthy tree annual RGRs among the six most abundant canopy species experiment-wide. Statistical analyses were performed in R (2019) using stats [45] and agricolae [46] statistical analysis packages. Statistical parameters associated with each analysis are presented in Appendix A Tables A1-A3. Raw data used in this analysis are available via the FoRTE data package [38]: https://github.com/ FoRTEXperiment/fortedata. Processed data products and statistical analysis specific to this study are available via: doi: 10.6084/m9.figshare.12442703.

Leaf Area Index
In the first growing season after disturbance, phloem disruption reduced mean end-of-season (Day of Year 213-215) LAI in the 85% disturbance severity treatment by 19% relative to the control and 45% severity treatment (Figure 2A; LSD = 0.62, d.f. = 21, α = 0.05). Mean end-of-season LAI was 3.19 in the control plots and the 45%, 2.79 in the 65%, and 2.57 in the 85% disturbance severity treatments, respectively. LAI did not differ significantly between bottom-up and top-down disturbance types ( Figure 2B; Raw data used in this analysis are available via the FoRTE data package [38]: https://github.com/FoRTEXperiment/fortedata. Processed data products and statistical analysis specific to this study are available via: doi: 10.6084/m9.figshare.12442703.

Leaf Area Index
In the first growing season after disturbance, phloem disruption reduced mean end-of-season (Day of Year 213-215) LAI in the 85% disturbance severity treatment by 19% relative to the control and 45% severity treatment (Figure 2A; LSD = 0.62, d.f. = 21, α = 0.05). Mean end-of-season LAI was 3.19 in the control plots and the 45%, 2.79 in the 65%, and 2.57 in the 85% disturbance severity treatments, respectively. LAI did not differ significantly between bottom-up and top-down disturbance types ( Figure 2B; LSD = 0.49 d.f. = 3, α = 0.05). Error bars are the SE of mean leaf area index (LAI) among treatment replicates. Nonoverlapping letters denote significance (α = 0.05) among disturbance severity treatments. Significant differences between disturbance types were not detected (n.s.).

Daily ANPPw and Disturbance Treatments
At the beginning of June 2019, two weeks after girdling, daily ANPPw in the control (0% disturbance severity) was significantly greater than that of the 65 and 85% disturbance severities with values of 13.89, 2.99, and 2.98 kg C ha −1 day −1 , respectively ( Figure   Error bars are the SE of mean leaf area index (LAI) among treatment replicates. Nonoverlapping letters denote significance (α = 0.05) among disturbance severity treatments. Significant differences between disturbance types were not detected (n.s.).

Daily ANPP w and Disturbance Treatments
At the beginning of June 2019, two weeks after girdling, daily ANPP w in the control (0% disturbance severity) was significantly greater than that of the 65 and 85% disturbance severities with values of 13.89, 2.99, and 2.98 kg C ha −1 day −1 , respectively ( Figure 3A, LSD = 8.56, d.f. = 336, α = 0.05). Within four weeks, however, daily ANPP w in the 65% disturbance severity reached 37.3 kg C ha −1 day −1 surpassing that of the control at 24.15 kg C ha −1 day −1 . After all disturbance severities reached a maximum in mid-July, daily ANPP w was significantly greater in both the 65 and 85% disturbance severity treatments relative to the control ( Figure

Relative Growth Rates of Girdled and Healthy Trees
When pooled across treatments to compare the RGRs of girdled and healthy trees, we found species growth responded differently to phloem disruption ( Figure 4). Among the six species examined, the RGR of girdled A. rubrum (t = 15.3, d.f. = 436, p < 0.001), A. saccharum (t = 11.29, d.f. = 145, p < 0.001), and P. grandidentata (t = 18.4, d.f. = 415, p < 0.001) were approximately double that of healthy trees of the same species ( Figure 4). The RGR of girdled P. strobus (t = 2.31, d.f. = 134, p = 0.02) was significantly, but more modestly, one-third greater than healthy P. strobus. In contrast, F. grandifolia (t = −0.4, d.f. = 345, p = 0.69) exhibited no response, while the RGR of girdled Q. rubra (t = −17.05, d.f. = 382, p < 0.001), a major component of the ecosystem, was two-thirds less than that of healthy Q. rubra (Figure 4).  . Mean daily upper canopy (>8 cm DBH) aboveground wood net primary production (ANPP w , ± 1 S.E.) by disturbance severity (A) and disturbance type (B) treatments in the first growing season following phloem-disrupting disturbance (vertical dashed line). Error bars are the SE of mean daily ANPP w among treatment replicates. Asterisks denote dates on which the control significantly (α = 0.05) differed from either the 65 or 85% disturbance severity. Significant differences between disturbance types were not detected (n.s.).

Figure 3.
Mean daily upper canopy (>8 cm DBH) aboveground wood net primary production (ANPPw, ± 1 S.E.) by disturbance severity (A) and disturbance type (B) treatments in the first growing season following phloem-disrupting disturbance (vertical dashed line). Error bars are the SE of mean daily ANPPw among treatment replicates. Asterisks denote dates on which the control significantly (α = 0.05) differed from either the 65 or 85% disturbance severity. Significant differences between disturbance types were not detected (n.s.).

Relative Growth Rates of Girdled and Healthy Trees
When pooled across treatments to compare the RGRs of girdled and healthy trees, we found species growth responded differently to phloem disruption ( Figure 4). Among the six species examined, the RGR of

Annual ANPP w and Disturbance Treatments
The aboveground growth of girdled trees sustained annual ANPP w in the first growing season after disturbance (Figure 5), resulting in no cumulative treatment effects on year-long production. Mean annual ANPP w was 2447 kg C ha −1 year −1 and did not differ significantly among disturbance severities or between disturbance types (Table A2). The contribution of girdled trees to annual ANPP w was roughly proportional to the targeted experimental disturbance severity levels, increasing significantly from 47 to 79 and 86% in the 45, 65 and 85% disturbance severity treatments, respectively ( Figure 5A; LSD = 576. 46, d.f. = 42, α = 0.05). Girdled trees in the bottom-up and top-down disturbance treatments accounted for 47% and 64% of annual ANPP w , respectively, with large diameter girdled trees contributing more to annual aboveground wood production in the latter treatment ( Figure 5B; LSD = 621.73, d.f. = 3, α = 0.05).

Annual ANPPw and Disturbance Treatments
The aboveground growth of girdled trees sustained annual ANPPw in the first growing season after disturbance (Figure 5), resulting in no cumulative treatment effects on year-long production. Mean annual ANPPw was 2447 kg C ha −1 year −1 and did not differ significantly among disturbance severities or between disturbance types (Table A2). The contribution of girdled trees to annual ANPPw was roughly proportional to the targeted experimental disturbance severity levels, increasing significantly from 47 to 79 and 86% in the 45, 65 and 85% disturbance severity treatments, respectively ( Figure 5A; LSD = 576. 46, d.f. = 42, α = 0.05). Girdled trees in the bottom-up and top-down disturbance treatments accounted for 47% and 64% of annual ANPPw, respectively, with large diameter girdled trees contributing more to annual aboveground wood production in the latter treatment ( Figure 5B; LSD = 621.73, d.f. = 3, α = 0.05).

Figure 5.
Contributions of girdled and healthy upper canopy (≥8 cm DBH) trees to annual aboveground wood net primary production (ANPPw, ± 1 S.E.) in the control and among disturbance severity (A) and disturbance type (B) treatments. Healthy and girdled refers to the fraction of ANPPw supplied by each within a disturbance severity (A) and disturbance type (B) category. Error bars are the SE of mean annual ANPPw among treatment replicates. While significant differences were not detected among treatments (n.s), the asterisk denotes that within disturbance types, girdled trees contributed more to ANPPw in the top-down relative to the bottom-up disturbance treatment (B; α = 0.05).
Among canopy strata, upper canopy trees were overwhelmingly the primary constituent of annual ANPPw, regardless of disturbance severity and type. Upper canopy trees contributed ~95% of total annual ANPPw in all disturbance treatments, with subcanopy and seedlings/saplings supplying the remainder of aboveground wood production ( Figure A1).

Successional Cohort Contributions to Annual ANPPw
The contribution of the early successional (Populus, Betula) species to annual ANPPw was not significantly different from that of the middle to late successional (Acer, Quercus, Pinus) species among disturbance severities (  Contributions of girdled and healthy upper canopy (≥8 cm DBH) trees to annual aboveground wood net primary production (ANPP w , ± 1 S.E.) in the control and among disturbance severity (A) and disturbance type (B) treatments. Healthy and girdled refers to the fraction of ANPP w supplied by each within a disturbance severity (A) and disturbance type (B) category. Error bars are the SE of mean annual ANPP w among treatment replicates. While significant differences were not detected among treatments (n.s), the asterisk denotes that within disturbance types, girdled trees contributed more to ANPP w in the top-down relative to the bottom-up disturbance treatment (B; α = 0.05).
Among canopy strata, upper canopy trees were overwhelmingly the primary constituent of annual ANPP w , regardless of disturbance severity and type. Upper canopy trees contributed~95% of total annual ANPP w in all disturbance treatments, with subcanopy and seedlings/saplings supplying the remainder of aboveground wood production ( Figure A1).

Successional Cohort Contributions to Annual ANPP w
The contribution of the early successional (Populus, Betula) species to annual ANPP w was not significantly different from that of the middle to late successional (Acer, Quercus, Pinus) species among disturbance severities (

Discussion
Contrary to our core hypothesis, the scaled aboveground wood production of girdled trees sustained stand-level ANPPw in the first year following disturbance, with little to no increase in healthy-tree compensatory growth. Anticipating a reduction during the first growing season in the aboveground wood production of girdled trees, we sampled ANPPw at a high frequency, expecting healthy tree growth to increase as limiting resource availability increased [24,[47][48][49]. A nearby landscape-scale study, in which a third of all canopy trees were stem girdled, demonstrated that nitrogen and light availability increased within two years after disturbance [24,31], stimulating healthy tree production [26,31]. In our experiment, marginal (<20%) reductions in LAI and the sustained or elevated growth of girdled trees suggest that resource reallocation from girdled to healthy trees was minimal in the first growing season. Consistent with our findings, Flower et al. (2013) found that emerald ash borer infected trees grew for two to five years following initial infestation [47,50], and experimental studies of individual trees report continued but often variable growth along with high survival rates in the first year following stem girdling [13,16,22]. However, we note that the response to phloem disruption varies among disturbance agents and host species, with mortality as rapid as one year following infestation by pine bark beetles [12,25]. In agreement with other studies, we found species' RGRs exhibited different initial responses to girdling [13,18,22], but our analysis demonstrated that these variable growth responses offset one another to stabilize ecosystem production even at the 85% disturbance severity level. Our results extend these findings to the whole ecosystem, demonstrating that the sustained growth of girdled trees was sufficient, despite wide variation among species, to maintain stand-scale ANPPw across a range of disturbance severities and types for the duration of the first growing season.
The mechanism supporting first-year aboveground wood production in girdled trees is likely sustained C fixation and the retention of photosynthate above the girdle. In healthy trees, recently fixed C is first allocated to the nearest sink before distribution to distal stem and roots sinks [19,51]. When phloem is severed, however, the transport of new photosynthate from the crown to the roots is eliminated entirely and C is retained in aboveground tissues [15]. In our study, higher late-season daily ANPPw in the middle and high (>65%) disturbance severity treatments suggest a progressive accumulation of C above the girdled phloem in most trees. However, immediate responses to phloem-disruption, as well as the timing of mortality, may vary among species because of differences in plant hormonal patterns [52,53], the depth and extent of tissue damage and repair [23,54] and feedback inhibition of leaf photosynthesis as carbohydrates accumulate [44,55,56]. For example, significantly lower contributions to ANPPw from girdled ring-porous Q. rubra relative to ring-diffuse Figure 6. The contribution of early and middle to late successional species cohorts from all canopy strata to annual aboveground wood net primary production (ANPP w , ± 1 S.E.) at each disturbance severity treatment. Error bars are the SE of mean annual ANPP w among treatment replicates. Significant differences (α = 0.05) between successional cohorts among disturbance severities were not detected (n.s.).

Discussion
Contrary to our core hypothesis, the scaled aboveground wood production of girdled trees sustained stand-level ANPP w in the first year following disturbance, with little to no increase in healthy-tree compensatory growth. Anticipating a reduction during the first growing season in the aboveground wood production of girdled trees, we sampled ANPP w at a high frequency, expecting healthy tree growth to increase as limiting resource availability increased [24,[47][48][49]. A nearby landscape-scale study, in which a third of all canopy trees were stem girdled, demonstrated that nitrogen and light availability increased within two years after disturbance [24,31], stimulating healthy tree production [26,31]. In our experiment, marginal (<20%) reductions in LAI and the sustained or elevated growth of girdled trees suggest that resource reallocation from girdled to healthy trees was minimal in the first growing season. Consistent with our findings, Flower et al. (2013) found that emerald ash borer infected trees grew for two to five years following initial infestation [47,50], and experimental studies of individual trees report continued but often variable growth along with high survival rates in the first year following stem girdling [13,16,22]. However, we note that the response to phloem disruption varies among disturbance agents and host species, with mortality as rapid as one year following infestation by pine bark beetles [12,25]. In agreement with other studies, we found species' RGRs exhibited different initial responses to girdling [13,18,22], but our analysis demonstrated that these variable growth responses offset one another to stabilize ecosystem production even at the 85% disturbance severity level. Our results extend these findings to the whole ecosystem, demonstrating that the sustained growth of girdled trees was sufficient, despite wide variation among species, to maintain stand-scale ANPP w across a range of disturbance severities and types for the duration of the first growing season.
The mechanism supporting first-year aboveground wood production in girdled trees is likely sustained C fixation and the retention of photosynthate above the girdle. In healthy trees, recently fixed C is first allocated to the nearest sink before distribution to distal stem and roots sinks [19,51]. When phloem is severed, however, the transport of new photosynthate from the crown to the roots is eliminated entirely and C is retained in aboveground tissues [15]. In our study, higher late-season daily ANPP w in the middle and high (>65%) disturbance severity treatments suggest a progressive accumulation of C above the girdled phloem in most trees. However, immediate responses to phloem-disruption, as well as the timing of mortality, may vary among species because of differences in plant hormonal patterns [52,53], the depth and extent of tissue damage and repair [23,54] and feedback inhibition of leaf photosynthesis as carbohydrates accumulate [44,55,56]. For example, significantly lower contributions to ANPP w from girdled ring-porous Q. rubra relative to ring-diffuse P. grandidentata, A. rubrum and A. saccharum, was likely caused by the disruption of the former's more abundant outer tracheid cells [54]. While continued, but temporary, aboveground growth of phloem-disrupted trees is well documented, studies do not agree on how stem girdling alters the partitioning of fixed C among aboveground tissues [13,15,16,57,58]. Further, the impending mortality and growth cessation of phloem-disrupted trees highlights the dynamic nature of C cycling responses to disturbance [59].
With girdled trees continuing to grow regardless of stem size class, top-down and bottom-up disturbance type treatments had no effect on first year ANPP w . As with naturally occurring disturbances, tree size and species identity at our site are intertwined [60] (Figure A2), with larger stems including both early successional P. grandidentata and later successional Q. rubrum. As a result, disentangling the independent effects of disturbance type from those of species identity is difficult. Our observation that girdled trees contribute more to annual ANPP w in the top-down rather than bottom-up disturbance (Figure 4) is consistent with observations that larger trees contribute more to stand-scale aboveground production than small trees [47,[61][62][63], a consequence of allometric scaling that occurs in most plants [64]. However, we anticipate differences in ANPP w to emerge between top-down and bottom-up disturbance treatments in subsequent years as the aboveground wood production of girdled trees declines and, conversely, increases in healthy trees.
Our findings do not support the hypothesis that early successional species contribute more to ANPP w than later successional species at higher disturbance severities in the first growing season after phloem-disrupting disturbance. At our site, the early successional cohort primarily consists of maturing aspen (Populus spp.) that thrive following severe, stand-replacing disturbances such as clear-cut harvesting and fire [65]. Stand-replacing disturbances abruptly increase the quantity of growth-limiting resources including light, nutrients and water, which may favor fast-growing, opportunistic pioneer species that aggressively compete for legacy resources [66][67][68]. Anticipating a more rapid first-year redistribution of growth-limiting resources from girdled to healthy trees, we expected early successional Populus to benefit more than middle and late successional species. Instead, the observed increase in Populus's RGR was similar to that of Acer spp. While stable growing season LAI in all but the highest severity disturbance treatment suggests that canopy light distribution changed minimally, increases in nutrient availability may occur more rapidly as leaf and root physiological activity diminishes in parallel with declining nutrient demand from girdled individuals [24,69,70]. Nonetheless, the continued growth of girdled trees suggests physiological decline was nominal and that resource redistribution from girdled to healthy trees, if it occurred, did not stimulate healthy-tree aboveground wood growth in the first year following phloem disruption. As girdled trees die, the canopy deteriorates, and light in the subcanopy becomes more abundant, early successional species may contribute more to aboveground wood production, particularly at high disturbance severities [31].
With little change in LAI during the first growing season following disturbance, we found that subcanopy, seedlings and saplings contributed minimally to ANPP w , even at high disturbance severities. In a nearby study, disturbance-driven changes in LAI were inversely related to subcanopy light availability and growth five years after stem girdling, with the subcanopy constituting nearly half of ANPP w at severities commensurate with our 65% disturbance severity treatment [31,71]. In comparison, we found that the subcanopy contributed~5% to annual ANPP w in the first year following disturbance, an amount comparable to the subcanopy contributions of local undisturbed closed canopy forests [71]. Our observations of limited subcanopy growth are consistent with concurrently measured leaf photosynthesis, chemistry and morphology that suggest no stimulation of subcanopy C fixation in the first growing season [30]. Nonetheless, the subcanopy contribution to total ANPP w is expected to increase over time as upper canopy LAI declines and subcanopy light availability increases [71][72][73][74].
Our finding of sustained ANPP w in the first growing season following stem girdling signals high initial resistance to disturbance, while highlighting the dynamic and variable nature of C cycling stability. Slowly unfolding disturbances, like phloem-disruption, provide time for the gradual reallocation of growth-limiting resources in a way that may be less ecologically feasible following more abrupt and severe disturbances. For example, clear-cut harvesting of a nearby forest rapidly increased nitrogen availability [70], but, in the absence of established vegetation, resulted in high leaching [75], which in turn prompted a century-long reduction in net primary production [34]. In contrast, a neighboring stem girdling manipulation displayed high nitrogen retention at disturbance severity levels comparable to our 45% severity treatment [24], sustaining longer term primary production stability [26,31] of this nitrogen-limited ecosystem [75]. Accordingly, initial C cycling responses to disturbance, in turn, may foreshadow longer-term changes in ecosystem function, with immediate responses in ecological processes linked to lasting changes in resource availability [76], microclimate [8] and plant competition [77,78]. While we observed minimal changes in ANPP w during the first year of our experiment, future stability will likely require the full and efficient transfer of growth-limiting resources to healthy vegetation once girdled trees begin to decline physiologically and their demand for nutrients, water, and light diminishes.
Our results have implications for the management of aboveground wood production following slow-acting disturbances. Because phloem-disrupting disturbances unfold slowly, real-time adaptive forest management strategies may effectively stabilize ecosystem functioning [79,80]. In contrast, adaptive management may be more limited following abrupt disturbances from stand-replacing fire, extreme weather or clear-cut harvesting [81]. In our study, the initial stand-scale stability of ANPP w suggests that managers have a window in which adaptive strategies could mitigate the long-term effects of slower-acting wood boring pests and pathogens. Such adaptive strategies might include stand-scale, growth-compensating measures such as the seeding or planting of subcanopy vegetation, along with production-efficiency increases associated with amended biodiversity and canopy structure manipulation [80][81][82][83]. Studies that document the efficacy of disturbance mitigation strategies are needed to develop broadly effective adaptive management approaches for the array of forest ecosystems and range of disturbance regimes observed in nature [79,[84][85][86].
We note several study limitations. First, allometries for estimating wood biomass were developed from the dimensions and wood densities of healthy trees and their use for girdled trees may introduce error if DBH-aboveground wood biomass relationships change following disturbance. Although plant-scale changes in allometry following phloem disruption have not been reported, an accumulation of C aboveground following phloem disruption is well documented [15,16]. Whether accumulated C is distributed evenly among aboveground structural and nonstructural C pools is unknown and could affect allometries and wood density, respectively [62,87,88]. While tapered swelling within 10 cm of the severed phloem has been reported for some species [21], we found moderately significant radial stem swelling in only one species (Acer rubrum) 30 cm above the girdled tissue, where we measured stem diameter. Moreover, while disproportionally high radial growth above the girdle could inflate stand-scaled production, our banding at multiple stem heights revealed the effects of A. rubrum swelling on ANPP w were not statistically significant (Table A4). Even so, allometric equations, when applied outside of their originating ecosystem, site and, in our case, experimental context, will yield estimates with greater unquantifiable uncertainty [89][90][91][92].

Conclusions
We conclude that a principal mechanism supporting aboveground wood production in the first growing season following a phloem-disrupting disturbance is the sustained growth of girdled trees. While the aboveground growth of Q. rubra declined in response to girdling, most species exhibited increased aboveground growth during the first year that, in aggregate, sustained aboveground stand-scale wood production. Counter to our hypothesis, the uninhibited growth of most girdled trees throughout the first growing season following girdling sufficiently sustained stand-scale ANPP w , even in the highest severity treatment affecting~85% of the woody vegetation. In addition to illuminating high frequency, first-year C cycling responses to phloem disruption, our findings lay the foundation for a comprehensive temporal understanding of the tree-to-ecosystem scale mechanisms supporting first-year wood production stability [59,93,94]. Finally, our results, when considered in the context of longer-term observations, indicate that the mechanisms supporting C cycling are dynamic, with phloem-disrupted tree growth inevitably giving way to other compensatory processes. Funding: Our work was funded by the National Science Foundation, Division of Environmental Biology, Award 1655095. We thank the University of Michigan Biological Station for logistical and technical support, and the use of research infrastructure.

Conflicts of Interest:
The authors declare no conflict of interest, and the funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Appendix A
Forests 2020, 11, x FOR PEER REVIEW 13 of 20 cycling are dynamic, with phloem-disrupted tree growth inevitably giving way to other compensatory processes.

Conflicts of Interest:
The authors declare no conflict of interest, and the funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.
Appendix A Figure A1. The contributions of upper canopy (≥8 cm diameter breast height, DBH), subcanopy (<8 cm DBH), and seedling/sapling (below DBH height) stems to annual aboveground wood net primary production (ANPPw) (+/− SE) in the first year following disturbance. Error bars are the SE of annual ANPPw among treatment replicates. Common letters ( a ) signify no significant (p > 0.05) difference in total (summed across canopy strata) annual ANPPw among disturbance severities. ANPPw ± 1 S.E.  cycling are dynamic, with phloem-disrupted tree growth inevitably giving way to other compensatory processes.

Conflicts of Interest:
The authors declare no conflict of interest, and the funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.
Appendix A Figure A1. The contributions of upper canopy (≥8 cm diameter breast height, DBH), subcanopy (<8 cm DBH), and seedling/sapling (below DBH height) stems to annual aboveground wood net primary production (ANPPw) (+/− SE) in the first year following disturbance. Error bars are the SE of annual ANPPw among treatment replicates. Common letters ( a ) signify no significant (p > 0.05) difference in total (summed across canopy strata) annual ANPPw among disturbance severities. ANPPw ± 1 S.E.   Table A1. Split-plot time-series analysis of variance (ANOVA) table for daily aboveground wood net primary production (ANPP w ), presented in Figure 3 of the manuscript. Week and disturbance type main effects and their interaction were tested against more conservative error terms to account for restricted randomization, whereas the main effect disturbance severity and its interaction with week were tested against the residual error. Asterisks indicate an interaction between two or more parameters.  Table A2. Split-plot analysis of variance (ANOVA) table for annual upper canopy (DBH ≥ 8 cm) aboveground wood net primary production (ANPP w ) presented in Figure 4 of the manuscript. Disturbance type and treatment (girdled/not-girdled) main effects and their interaction terms were tested against more conservative error terms to account for restricted randomization, whereas the main effect disturbance severity and its interaction with treatment were tested against the residual error. Asterisks indicate an interaction between two or more parameters.  Table A3. Split-plot analysis of variance (ANOVA) table for annual aboveground wood net primary production (ANPP w ), presented in Figure 5 of the manuscript. Disturbance type and successional cohort (cohort) main effects and their interaction were tested against more conservative error terms to account for restricted randomization, whereas the main effect disturbance severity and its interaction with cohort were tested against the residual error. Asterisks indicate an interaction between two or more parameters.  Table A4. Annual upper canopy (DBH ≥ 8 cm) aboveground wood net primary production (ANPP w ) values unadjusted for swelling and adjusted for Acer rubrum DBH swelling (30 cm) above girdled tissue. Dendrometer bands (N = ≥3 per species) were fitted to stems at two heights 50 cm apart and their diameter increments recorded concurrently on five occasions between July 2019 and November 2019. Among the five species examined, only Acer rubrum exhibited significantly greater diameter growth taper in girdled trees (p = 0.06). To evaluate the sensitivity of annual ANPP w to Acer rubrum swelling, we scaled production separately including (unadjusted) and excluding (adjusted) Acer rubrum contributions to annual ANPP w . We found no significant differences between unadjusted and adjusted annual ANPP w estimates (unpaired t-tests; α = 0.05), nor did the unadjusted or adjusted ANPP w estimates affect the outcome of pairwise treatment comparisons (least squares differences, pairwise comparisons; α = 0.1).