Forest Regeneration Following Emerald Ash Borer ( Agrilus planipennis Fairemaire) Enhances Mesophication in Eastern Hardwood Forests

: Emerald ash borer (EAB, Agrilus planipennis Fairemaire) is a phloem-feeding beetle that was introduced into North America in the late 20th century and is causing widespread mortality of native ash ( Fraxinus ) species. The loss of an entire genus from the forest ﬂora is a substantial disturbance, but effects vary because of differences in Fraxinus dominance and remaining vegetation. At three sites near the center of the North American EAB range, we investigated the impacts of Fraxinus mortality on recruitment of woody and non-native vegetation in 14 permanent plots from 2012 to 2017. We used the change in relative Fraxinus basal area to determine the impact of EAB on density of woody species and non-native vegetation less than 2.5 cm diameter at breast height (dbh). Changes in canopy cover were not correlated with loss of Fraxinus from the overstory, and only the density of shade-tolerant shrubs and saplings increased with Fraxinus mortality. Both native and non-native shrub species increased in density at sites where they were present before EAB, but no new invasions were detected following Fraxinus mortality. These shifts in understory vegetation indicate that Fraxinus mortality enhances the rate of succession to shade-tolerant species.


Introduction
In the nearly two decades since emerald ash borer (EAB, Agrilus planipennis Fairemaire) was first detected in North America [1], the response has evolved from preventing or mitigating damage to Fraxinus [2][3][4] to understanding how the introduced insect might impact other species and ecosystem function in forest communities [5][6][7]. Forest plant communities of eastern North America, like any system, are dynamic and subject to change, and EAB's impact on Fraxinus is not exceptional. By inducing Fraxinus mortality, EAB immediately reduces forest stand basal area and releases resources, which subsequently become available to other species [8][9][10].
As a component of mesophytic forests of eastern North America, particularly those within Eastern US Lowland Forest Region-Upland Forest Types [11], Fraxinus species might well be considered a minor component of the system, represented primarily by F. americana L. and F. pennsylvanica Marshall [12]. But these species have a wide distribution, and despite their relatively low abundance in this forest type [12], the loss of Fraxinus by EAB has the potential to drive changes in plant communities throughout the region, similar in respect to previous introductions that changed forest composition on a broad scale. Chestnut blight (Cryphonectria parasitica (Murrill) Barr) led to the widespread loss of Castanea dentata (Marshall) Borkh. throughout its native North American range, particularly in the Appalachian region [13], and Dutch elm disease (Ophiostoma ulmi Buisman (Melin from 847 to 1132 mm, and mean annual temperature varies from 9.6 to 11.8 • C (Table 1). Soils are predominantly finer textured, from silty loam to silty clay loam [32].

Data Collection
We used the existing infrastructure of the Permanent Forest Plot Project (PFPP), one of the projects associated with the Ecological Research as Education Network (EREN; http://erenweb.org/). This project was launched in 2012 to establish a set of permanent research plots to address questions related to tree biomass, carbon accumulation, invasive species, and disturbance patterns across a range of sites and ecoregions [34]. Participants establish at least 1 permanent 400-m 2 plot in a forested area, identify and measure the diameter of all trees (>2.5 cm diameter at breast height (dbh), or 1.37 m above the ground), identify and tally all saplings (0-2.5 cm dbh, but at least 1.37 m tall) in 3 randomly selected subplots (25 m 2 ), and characterize their site and plot using a standard protocol [34]. PFPP data are entered in a secure online database accessible to all PFPP participants.
We also scaffolded 4 variables onto PFPP to evaluate the impact of EAB on Fraxinus trees. Tree damage indicates whether the tree is broken, uprooted, or leaned on by another tree (including Fraxinus); this variable was incorporated into the PFPP protocol [34]. Ash rating [35] and ash tree breakup [36] indicate the level of decline, mortality, and breaking down of Fraxinus trees. We also counted the number of EAB exit holes observed between 1.25 and 1.75 m above the ground as a physical indicator that EAB is associated with the Fraxinus decline. These 3 variables are specific to the EAB Impacts Study [37], another EREN project.
To address questions related to the successional dynamics of forests, we built variables to characterize the understory plant community in the Complementary Vegetation (cVeg) Survey [38]. Canopy cover was measured using a concave spherical densiometer (Model C, Robert E. Lemmon, Rapid City, SD, USA), while all woody plants between 0.3 and 1.37 m tall (i.e., shrubs and saplings) were identified and tallied within each of the three 25-m 2 subplots used in the PFPP. In addition, all woody seedlings (<0.3 m tall) and non-woody plants were identified and tallied within 1-m 2 miniplots in each of the subplots. Woody species were classified according to their shade tolerance [39,40].

Analyses
Given our hypothesis that the loss of Fraxinus is driving a shift in community composition toward more mesophytic, or shade-tolerant, species, as well as increasing the prevalence of invasive plant species, we used the preinvasion relative basal area of Fraxinus as the independent variable. This variable was calculated as the proportion of cumulative basal area (BA = π × (0.5 × dbh) 2 ) of 3 Fraxinus species (F. americana, F. pennsylvanica, F. quadrangulata Michx.) relative to all other trees prior to (BW and WJ sites) or at (UF sites) the time of EAB detection. The variables of interest, such as annual changes in woody seedling density, invasive plant density, and sapling and shrub density, partitioned by shade tolerance [39,40], were calculated as the annual change in density from the first to last sampling year and divided by the number of years of sampling, whether we had 2 (BW) or 5 (UF, WJ) years of data. Data from miniplots (woody seedling and forb) and subplots (shrubs and saplings) were pooled by plot, thus plot became the observational unit. Each of these dependent variables was then linearly regressed onto the relative basal area of Fraxinus. All analyses were conducted using IBM SPSS Statistics v.24 (IBM Corp., North Castle, NY, USA).

Results
To determine whether we could detect changes in light resources at the forest floor, we regressed change in canopy cover on relative Fraxinus basal area prior to EAB invasion. Across all four sites, we found no significant relationship between canopy cover and relative Fraxinus dominance (p = 0.46), thus no evidence that light resources increased more in sites with greater dominance of Fraxinus ( Figure 3).  The relative basal area of Fraxinus has not significantly affected the density of woody seedlings across sites (p > 0.18), whether with combined species or when partitioned by shade tolerance ( Figure  4), yet some patterns are emerging. For example, shade-tolerant species increased in density at a higher rate in sites with more Fraxinus trees. Seedling density for F. americana generally declined across all sites, yet F. pennsylvanica seedlings generally increased in density at both UF sites (in supplementary file Table S1). Across UF and WJ sites, the density of Parthenocissus quinquefolia (L.) Planch. seedlings increased since EAB invasion. Other species, including the non-native Rosa multiflora Thunb. experienced both increased and decreased density within sites (Table S1). The relative basal area of Fraxinus has not significantly affected the density of woody seedlings across sites (p > 0.18), whether with combined species or when partitioned by shade tolerance (Figure 4), yet some patterns are emerging. For example, shade-tolerant species increased in density at a higher rate in sites with more Fraxinus trees. Seedling density for F. americana generally declined across all sites, yet F. pennsylvanica seedlings generally increased in density at both UF sites (in supplementary file Table S1). Across UF and WJ sites, the density of Parthenocissus quinquefolia (L.) Planch. seedlings increased since EAB invasion. Other species, including the non-native Rosa multiflora Thunb. experienced both increased and decreased density within sites (Table S1).  The relative basal area of Fraxinus has not significantly affected the density of woody seedlings across sites (p > 0.18), whether with combined species or when partitioned by shade tolerance ( Figure  4), yet some patterns are emerging. For example, shade-tolerant species increased in density at a higher rate in sites with more Fraxinus trees. Seedling density for F. americana generally declined across all sites, yet F. pennsylvanica seedlings generally increased in density at both UF sites (in supplementary file Table S1). Across UF and WJ sites, the density of Parthenocissus quinquefolia (L.) Planch. seedlings increased since EAB invasion. Other species, including the non-native Rosa multiflora Thunb. experienced both increased and decreased density within sites (Table S1).  While total shrub density (Figure 5a), including shade-intolerant species (Figure 5c), did not respond to changes in relative basal area of Fraxinus, shade-tolerant shrubs significantly (p = 0.005) increased in density (Figure 5b). This response appears to be driven by individual plots at the BW (Rocky-1) and UF (Olive-2) sites, both of which had higher losses of Fraxinus than other plots ( Figure  2); two native shrubs, Lindera benzoin L. and Asimina triloba (L.) Dunal, respectively, substantially increased in these plots, but the non-native Rosa multiflora also increased in density over this same period at UF (Olive-2) (in supplementary file Table S2). Saplings of Fraxinus spp. generally increased at UF (Rieck) and WJ (AFS), as did the native Celtis occidentalis at UF (Rieck) and non-native Celastrus orbiculatus Thunb. and native Rubus spp. at WJ (AFS). However, Lindera benzoin generally decreased in density at WJ (AFS), while the non-native Lonicera spp. and Rosa multiflora Thunb. increased in the UF (Olive-1) plot that experienced widespread canopy tree loss from the June 2012 North American derecho (Table S2).  While total shrub density (Figure 5a), including shade-intolerant species (Figure 5c), did not respond to changes in relative basal area of Fraxinus, shade-tolerant shrubs significantly (p = 0.005) increased in density (Figure 5b). This response appears to be driven by individual plots at the BW (Rocky-1) and UF (Olive-2) sites, both of which had higher losses of Fraxinus than other plots ( Figure 2); two native shrubs, Lindera benzoin L. and Asimina triloba (L.) Dunal, respectively, substantially increased in these plots, but the non-native Rosa multiflora also increased in density over this same period at UF (Olive-2) (in supplementary file Table S2). Saplings of Fraxinus spp. generally increased at UF (Rieck) and WJ (AFS), as did the native Celtis occidentalis at UF (Rieck) and non-native Celastrus orbiculatus Thunb. and native Rubus spp. at WJ (AFS). However, Lindera benzoin generally decreased in density at WJ (AFS), while the non-native Lonicera spp. and Rosa multiflora Thunb. increased in the UF (Olive-1) plot that experienced widespread canopy tree loss from the June 2012 North American derecho (Table S2). Relative basal area of Fraxinus was not a significant predictor of density of non-native vegetation in the ground layer or the shrub and sapling layer in 2017 ( Figure 6). Only two plots, one each at WJ (AFS) and UF (Olive), saw above-average increases in non-native species density, yet both sites had relatively low relative amounts of Fraxinus basal area. And though three non-native species, Celastrus orbiculatus Thunb., Lonicera maackii (Rupr.) Maxim. and Rosa multiflora, increased in density in the shrub layer within sites during the period of study (Table S2), the overall change in non-native species density does not show a relationship to the decrease in relative Fraxinus basal area.  Relative basal area of Fraxinus was not a significant predictor of density of non-native vegetation in the ground layer or the shrub and sapling layer in 2017 ( Figure 6). Only two plots, one each at WJ (AFS) and UF (Olive), saw above-average increases in non-native species density, yet both sites had relatively low relative amounts of Fraxinus basal area. And though three non-native species, Celastrus orbiculatus Thunb., Lonicera maackii (Rupr.) Maxim. and Rosa multiflora, increased in density in the shrub layer within sites during the period of study (Table S2), the overall change in non-native species density does not show a relationship to the decrease in relative Fraxinus basal area.

Discussion
The loss of Fraxinus from mesophytic North American hardwood forests and the subsequent release of resources appear to have little influence on the distribution and abundance of forest regeneration at the seedling and sapling level. Furthermore, we found no evidence that this loss alters the process of mesophication in the same manner as acute, severe disturbances that typically impact canopy cover.

Canopy Cover
Light is the primary resource expected to change at the ground layer as a result of the death of large overstory Fraxinus trees [41], yet our evidence suggests that the protracted decline and ultimate death of Fraxinus individuals does not create conditions that substantially alter the light regime on the forest floor, as measured by our canopy cover index. Although canopy cover decreased at a rate of 0%-3% annually, there was no detectable relationship to Fraxinus mortality ( Figure 3). In order to meet the physiological demands of shade-intolerant species, like Quercus and Carya, light levels need to reach 25%-50% full sunlight [42], which is often reached with more severe disturbances, like shelterwood harvesting [43].
Research on canopy trees in response to EAB-induced Fraxinus mortality shows that existing canopy trees respond positively. Gaps concurrently fill with non-Fraxinus species [23], likely influencing the lack of substantial change in annual canopy cover over time. Recent evidence using LIDAR (Light Detection and Ranging) indicates that single-tree death in northern hardwood forests is not sufficient to generate canopy gaps where the subcanopy is well developed [44], suggesting that the gradual canopy opening that occurs with EAB-induced Fraxinus mortality is less likely to enrich the ground layer with light than it is to promote the growth of existing subcanopy trees.
Indeed, in northwest Ohio, the relative growth rates of the shade-tolerant species Acer saccharum and Ulmus americana were faster than other species in the forest canopy following Fraxinus mortality [8], and smaller trees exhibited faster growth rates than larger ones, suggesting that subcanopy individuals benefit more than the co-dominant overstory trees. Trees remaining in the canopy of our forest sites include mesophytic species, typically from Acer, Prunus, Ulmus, and Celtis (Table 1), which supports our hypothesis that Fraxinus loss will increase the speed at which forests are dominated by more shade-tolerant species.

Discussion
The loss of Fraxinus from mesophytic North American hardwood forests and the subsequent release of resources appear to have little influence on the distribution and abundance of forest regeneration at the seedling and sapling level. Furthermore, we found no evidence that this loss alters the process of mesophication in the same manner as acute, severe disturbances that typically impact canopy cover.

Canopy Cover
Light is the primary resource expected to change at the ground layer as a result of the death of large overstory Fraxinus trees [41], yet our evidence suggests that the protracted decline and ultimate death of Fraxinus individuals does not create conditions that substantially alter the light regime on the forest floor, as measured by our canopy cover index. Although canopy cover decreased at a rate of 0%-3% annually, there was no detectable relationship to Fraxinus mortality ( Figure 3). In order to meet the physiological demands of shade-intolerant species, like Quercus and Carya, light levels need to reach 25%-50% full sunlight [42], which is often reached with more severe disturbances, like shelterwood harvesting [43].
Research on canopy trees in response to EAB-induced Fraxinus mortality shows that existing canopy trees respond positively. Gaps concurrently fill with non-Fraxinus species [23], likely influencing the lack of substantial change in annual canopy cover over time. Recent evidence using LIDAR (Light Detection and Ranging) indicates that single-tree death in northern hardwood forests is not sufficient to generate canopy gaps where the subcanopy is well developed [44], suggesting that the gradual canopy opening that occurs with EAB-induced Fraxinus mortality is less likely to enrich the ground layer with light than it is to promote the growth of existing subcanopy trees.
Indeed, in northwest Ohio, the relative growth rates of the shade-tolerant species Acer saccharum and Ulmus americana were faster than other species in the forest canopy following Fraxinus mortality [8], and smaller trees exhibited faster growth rates than larger ones, suggesting that subcanopy individuals benefit more than the co-dominant overstory trees. Trees remaining in the canopy of our forest sites include mesophytic species, typically from Acer, Prunus, Ulmus, and Celtis (Table 1), which supports our hypothesis that Fraxinus loss will increase the speed at which forests are dominated by more shade-tolerant species.

Understory Composition and Density
The composition of woody species in the seedling layer is relatively stable following the loss of Fraxinus to EAB. We detected no statistically significant relationship between the loss of Fraxinus and the change in seedling density (Figure 4). In regard to individual species, there is a large amount of variability among sites and plots. Species with the largest, most consistent increases over time include the native vine Parthenocissus quinquefolia, while the non-native shrub Rosa multiflora increased or decreased across sites, indicating a lack of relationship with Fraxinus loss. Interestingly, the number of Fraxinus seedlings increased at both UF (Olive) plots, where overstory Fraxinus trees had succumbed to EAB beginning in 2012 (Table S1). Recent evidence suggests that Fraxinus seedbanks do not persist beyond the life of the tree [45], indicating that an off-site seed source persists for this site.
The shrub and sapling layer was more responsive to Fraxinus mortality than the seedling layer. As the amount of Fraxinus basal area increased in plots, the density of shade-tolerant species increased (r 2 = 0.50, p = 0.005). This was in contrast to the lack of association detected between Fraxinus basal area and the change in shade-intolerant species density (r 2 = 0.003, p = 0.853). The increase in shade-tolerant species density coincides with evidence of established trees in the mid-canopy responding to the resources made available through the mortality of Fraxinus individuals. The response of individual species varied across sites, but those with increases at more than one site include shade-tolerant Lindera benzoin and Asimina triloba, both native understory species.
Our findings are not unusual, as shrub and sapling density has been shown to increase in response to species-specific canopy dieback elsewhere. In Europe, fungal-induced Fraxinus excelsior L. dieback promoted the growth of widely distributed shade-tolerant shrubs and understory trees, including Corylus avellana L., Prunus padus L., and Lonicera xylosterum L. [46]. In North America, after Dutch elm disease induced Ulmus decline, the intermediate-and shade-tolerant shrubs Alnus rugosa (L.) Moench, Viburnum recognitum Fern., and Cornus stolonifera L. in wetland habitats responded more positively in gaps created by the death of Ulmus stems in the overstory than seedlings of other canopy trees [47]. In southeastern Wisconsin, species composition and density of woody seedlings were not associated with density of dead Ulmus, yet the density of shrubs, especially Ribes spp., Rubus spp., and Cornus racemosa Lam., was associated with Ulmus mortality [48].

Invasive Species
No new invasive species were detected in the research plots. For plots with existing invasive plants, some became more numerous, including Lonicera maackii, Rosa multiflora, and Celastrus orbiculatus, but because of variability among plots, we were unable to identify Fraxinus mortality as the primary cause of the increase at our sites. Prior to EAB invasion, one plot at the UF (Olive) site had large numbers of L. maackii and R. multiflora; these species increased in density over time. This particular site, however, was also strongly influenced by a storm that caused significant loss of the overstory in 2012. The confounding disturbances of storm damage and EAB-induced Fraxinus mortality cannot be distinguished. Recent evidence suggests that the amount of Fraxinus, and thus the amount of light increasing beneath the canopy, influences the response of existing L. maackii; in a comprehensive study of population-level responses to EAB, existing L. maackii individuals had increased stem diameter following Fraxinus mortality, while the number of individual plants, amount of fruiting, and height of the shrubs were not impacted [9].
Additionally, in three plots in Pennsylvania, Celastrus orbiculatus had higher annual increases in density than other species following EAB, but because of losses to this species in other plots at the same site and an overall lack of relationship between Fraxinus basal area and shrub and sapling density (Figure 6b; r 2 = 0.001, p = 0.955), the increase in this species was not likely caused by the loss of Fraxinus.

Conclusions
Our evidence shows that the resources released through Fraxinus mortality are insufficient to modify the process of mesophication in mesophytic forests, but instead generate changes that resemble those observed with small gaps. The primary beneficiaries of released resources include individuals already existing in the understory canopy, among them small shade-tolerant trees, saplings, and shrubs typical of mature mesophytic forests, conforming to what is observed in forest gap dynamics [8,[49][50][51][52]. This enhanced mesophication resulting from the loss of Fraxinus has allowed shade-tolerant species, including those restricted to understory canopy positions, like Asimina triloba (L.) Dunal and Lindera benzoin L., to increase in the shrub and sapling layer at faster rates than shade-intolerant species ( Figure 5). As mesophication has been noted throughout much of the eastern North American forest complex [28,[53][54][55], we would expect to see similar patterns throughout the region where F. pennsylvanica and F. americana are the primary Fraxinus species represented in the canopy.
Importantly, some of our sites are in regions with relatively high amounts of Fraxinus compared to other mesophytic forest regions of eastern North America [11]. Our research primarily includes sites with relative Fraxinus basal area within the range of 5-25%, and we find no compelling evidence that the loss of trees at our sites impacts regeneration and recruitment. Where Fraxinus is a smaller component of the forest tree community, we expect that, like our sites, these areas will not see dramatic shifts in seedling and shrub composition resulting from the loss of Fraxinus in the overstory.
With the exception of Fraxinus, EAB is not a disturbance that influences the abundance of woody species regeneration and subsequent recruitment in sites where Fraxinus is a minor component of the canopy vegetation. The five species of Fraxinus most common to eastern hardwood forests, F. americana, F. pennsylvanica, F. nigra, F. profunda Bush, and F. quadrangulata, have been classified as critically endangered and on the brink of extinction by the International Union for Conservation of Nature (IUCN) [56][57][58][59][60]. Conservation of the genus is important, but the consequence of losing the species has little influence on immediate future successional pathways of woody vegetation in the forests where Fraxinus is a minor component.
Author Contributions: B.J.D. and J.S.K. developed the research questions, conceived the design for the EAB Impacts Study and Complementary Vegetation (cVeg) Survey, collected the data, performed the analyses, and wrote the manuscript. Both B.J.D. and J.S.K. provided edits and comments during the drafting of the paper. We are grateful to our home institutions for funding and for supporting travel to work together, and we are thankful for the contributions of data, discussions, and analyses of changes in the invasive plant species by Kathryn Flinn (Baldwin-Wallace University). For access to field sites, we thank Cleveland Metroparks (Rocky River), University of Findlay (Olive Street, Rieck Center), and the Abernathy family (Abernathy Field Station). We appreciate the online site data gained from the PFPP database. We also thank the myriad undergraduate students and two high school students who contributed to data collection, and the reviewers who improved this manuscript with insightful comments and suggestions.

Conflicts of Interest:
The authors declare no conflict of interest.