Forest disturbance and arthropods: Small-scale canopy gaps drive invertebrate community structure and composition

In forest ecosystems, disturbances that cause tree mortality create canopy gaps, increase growth of understory vegetation, and alter the abiotic environment. These impacts may have interacting effects on populations of ground-dwelling invertebrates that regulate ecological processes such as decomposition and nutrient cycling. A manipulative experiment was designed to decouple effects of simultaneous disturbances to the forest canopy and ground-level vegetation to understand their individual and combined impacts on ground-dwelling invertebrate communities. We quantified invertebrate abundance, richness, diversity, and community composition via pitfall traps in response to a factorial combination of two disturbance treatments: canopy gap formation via girdling and understory vegetation removal. Formation of gaps was the primary driver of changes in invertebrate community structure, increasing activityabundance and taxonomic richness, while understory removal had smaller effects. Families of Collembola and Diplopoda, as well as some families of Coleoptera, increased in combined canopy and understory disturbance treatments, whereas Curculionidae and Nitidulidae were more abundant in undisturbed forest. Gaps increased light availability, height and cover of understory vegetation, and soil moisture levels, and decreased depth and cover of leaf litter compared to undisturbed forest. Decoupling of canopy and understory vegetation disturbances revealed gap formation as an important short-term driver of grounddwelling invertebrate community structure and composition. Our findings increase understanding of how ground-dwelling invertebrate communities respond to disturbance and inform sustainable management of forest ecosystems to foster biodiversity and resilience.


INTRODUCTION
Disturbances alter energy and nutrient flow, habitat structure, and species composition in ecosystems (White and Pickett 1985), which increases spatial and temporal heterogeneity at multiple spatial scales (Oliver andLarson 1996, Frelich 2002). These events shape fluctuations in community dynamics and ecosystem processes over time (Runkle 1985, Schowalter 1985, Pickett et al. 1989. In forest ecosystems, disturbances can cause tree mortality and formation of canopy gaps. Gaps alter the forest floor environment by increasing light availability, altering soil temperature and moisture regimes, reducing leaf litter moisture and depth, and changing understory vegetation composition and structure (Collins et al. 1985, Gray et al. 2002, Ishizuka et al. 2002, Fahey and Puettmann 2007, Perry and Herms 2016b. However, increased light availability on the forest floor stimulates growth of understory vegetation, seedlings, saplings, and advanced regeneration (Oliver andLarson 1996, Nelson et al. 2008), which then can mediate the initial abiotic conditions induced by canopy gaps (Shure and Phillips 1991). Therefore, it is probable that changes to the canopy and understory following disturbance interact to impact the abundance and diversity of grounddwelling invertebrates (Shure and Phillips 1991, Bouget and Duelli 2004, Gandhi et al. 2008, Perry and Herms 2016a.
Ground-dwelling invertebrate communities are useful for detecting and characterizing the effects of multiple, interacting disturbances on forest ecosystems (Moldenke et al. 2000) because they respond quickly to changes in habitat complexity and microclimate conditions such as soil moisture (Levings and Windsor 1984), coarse woody debris (CWD; Ulyshen and Hanula 2009), leaf litter depth (Koivula et al. 1999), and vegetation cover and structure (Shure andPhillips 1991, Pakeman andStockan 2014). Spiders (Araneae), springtails (Collembola), ground beetles (Carabidae), and ants (Formicidae) are common taxa that have been used as biological indicators to detect and monitor changes in the abiotic environment in response to disturbance (Folgarait 1998, Rainio and Niemel€ a 2003, Pearce and Venier 2006, Greenslade 2007. Effects of perturbations on ground-dwelling invertebrates have implications for ecosystem services such as decomposition, nutrient cycling, and maintenance of soil structure (Brussaard 1997, Ruiter et al. 2002, Lavelle et al. 2006. Experiments that decouple impacts of interacting disturbances are often unfeasible, but necessary to evaluate the effect size of each factor individually. Studies that decoupled the effects of canopy gaps and accumulation of woody debris (Richardson et al. 2010, Perry and Herms 2016a and microhabitats created by tip-up mounds from fallen trees ) have identified canopy gaps as a major driver of ground-dwelling invertebrate community structure and composition. Because increased light availability from the formation of canopy gaps stimulates understory growth and alters vegetation composition (Collins et al. 1985, Oliver andLarson 1996), abiotic conditions on the forest floor may change following the response of the understory (Royo and Carson 2006).
The objective of this study was to investigate the individual and interacting effects of disturbance to the forest canopy and ground-level understory vegetation on ground-dwelling invertebrate communities by decoupling these two factors via a manipulative experiment. We predicted that (1) formation of canopy gaps would be the dominant environmental factor structuring ground-dwelling invertebrate communities; (2) invertebrate activity-abundance and diversity would decrease in response to combined canopy gap formation and ground-level understory vegetation removal; and (3) presence of ground-level understory vegetation in canopy gaps would increase invertebrate activity-abundance and diversity by buffering variability in environmental changes on the forest floor.

Study site
Research was conducted at Powdermill Nature Reserve (PNR; 40°09 0 S, 79°16 0 W) in Rector, Westmoreland County, Pennsylvania, in the Laurel Highlands. Powdermill Nature Reserve has approximately 900 ha of natural habitat across an elevation range of 392-647 m that is largely temperate deciduous forest composed of mesophytic species. Powdermill Nature Reserve was established as a nature reserve and the field research station for the Carnegie Museum of Natural History in 1956 and has been largely unmanaged since. Prior to this, the region experienced several major anthropogenic disturbances, including logging in the 19th century, agricultural production until the early-to mid-20th century, and then some areas of the reserve were surface mined for coal in the 1940s.

Experimental design
A manipulative experiment was established in the forests of PNR with two disturbance treatments: the presence/absence of canopy gaps and the presence/absence of ground-level understory vegetation removal, resulting in four treatment combinations including (1) closed canopy/understory present (undisturbed control); (2) closed canopy/understory removed; (3) canopy gap/understory present; and (4) canopy gap/understory removed. The four treatment combinations were distributed randomly among 24 quadrats 30 9 30 m in size (the experimental unit of replication) such that each treatment combination was replicated six times. Quadrats were identified in 2013 and treatments were implemented once in June 2014, such that disturbance intensity was manipulated rather than frequency. Eight undisturbed reference quadrats (30 9 30 m) were included for additional undisturbed controls (closed canopy/understory present).
To implement the canopy gap treatment, 8-12 dominant or codominant trees were girdled on 5-6 June 2014 to create gaps in the forest canopy (Appendix S1: Fig. S1). To girdle the trees, a chainsaw was used to cut a 2 cm wide ring (6-12 cm deep) around the entire circumference of the trunk about 1.5 m above the ground. A second ring was cut approximately 20-30 cm below the first. Across all canopy gap treatments, 117 trees were girdled, of which the most common were the diffuse-porous species red maple (Acer rebrum L.), sugar maple (Acer saccharum Marsh.), and tulip poplar (L. tulipifera L.). All girdled tree species listed in Appendix S1: Fig. S3).
The number of trees girdled in each canopy gap treatment varied depending on the density and size of trees in the quadrat and was determined by visually estimating the amount of cumulative canopy removal required to achieve the targeted gap size. Canopy gap size and the health of girdled trees were evaluated to assess changes in the canopy gap treatment during the study. Percentage canopy openness was quantified in all treatments once in July 2013 before treatments were implemented and monthly during the summers of 2014 and 2015. Degree of canopy dieback was visually assessed for each girdled tree biweekly and monthly during the summers of 2014 and 2015, respectively, using a 1-5 health rating scale (Smith et al. 2015), where one corresponded to a healthy tree with a full canopy, five corresponded to a dead tree, and 2-4 represented stages of increasing canopy thinning. All girdled trees were categorized as either diffuse-or ring-porous, as ring-porous trees were expected to decline faster due to their thinner functional xylem.
To implement the understory removal treatment, mechanical weed trimmers (KM 110 R KombiMotor, Stihl; Virginia Beach, Virginia, USA) and loppers were used to remove all understory herbaceous plants, woody shrubs, seedling, and saplings (≤2 cm stem diameter) within the 30 9 30 m experimental quadrats (Appendix S1: Fig. S2). Herbaceous vegetation residue remained on the forest floor in the quadrats, while all woody shrubs and saplings were removed and deposited in the surrounding forest away from the treatments. Understory vegetation was removed on 5-7 June 2014. Height and percentage cover of understory vegetation were quantified in two randomly selected 1 m 2 areas in all treatments once in 2013 before treatments were implemented and monthly during the summers of 2014 and 2015 to assess changes in the understory during the study.
passive method commonly used to collect ground-dwelling invertebrates Niemel€ a 1994, Yi et al. 2012). Because this method tends to collect more active invertebrates, abundance of invertebrates is presented as activityabundance (individuals per trap) and results should be interpreted accordingly. One trap was installed 5 m from the center of each experimental quadrat in a randomly selected cardinal direction. Barrier pitfall traps consisted of two pairs of plastic cups (each pair consisted of an inner 500-mL and outer 1-L plastic cup) dug into the ground such that the rim of cups was flush with the soil surface and connected by garden edging (Eco Edge, Suncast, Batavia, Illinois, USA) 1 m in length. The garden edging was used as a barrier to increase trap catch (Durkis and Reeves 1982). The inner cup contained 4 cm of propylene glycol (recreational vehicle and marine antifreeze, Peak Company Old World Industries, Clear Lake, Texas, USA) along with a few drops of detergent to facilitate collection. Masonite board (100 cm 2 ) was placed 3 cm above each pair of cups to prevent flooding from rainwater, and steel hardware cloth was secured over each pair with 30-cm stakes to limit animal disturbance.

Forest floor environment
Environmental conditions were quantified to assess the effects of the disturbance treatments on the forest floor microclimate. Percentage cover of leaf litter, bare ground, fine woody debris (FWD; <10 cm in diameter at the large end), CWD (≥10 cm in diameter at the large end), and rocks were assessed in two randomly selected 1 m 2 areas surrounding each pitfall trap once in 2013 and monthly during the summers of 2014 and 2015. Environmental conditions were quantified on 30-31 July in 2013; on 17-19 June, 29-31 July, and 26-27 August in 2014; and on 9-10 June, 7-8 July, and 5-6 August in 2015. Leaf litter depth was measured on 29-31 July 2014. Volume of downed CWD was estimated along 30-m transects within each treatment in 2014 using the line intercept method described by Harmon and Sexton (1996). Downed CWD along the transects was categorized according to decay class using a 1-5 decay stage scale adapted from Pyle and Brown (1999) and Woodall and Williams (2005). Soil temperature and moisture were measured adjacent to pitfall traps with a Durac Bi-Metallic Dial Thermometer and Dynamax TH 2 O Portable Soil Moisture Meter and Theta Probe ML2x Sensor, respectively. Measurements were collected biweekly when trap samples were collected. Three readings were averaged together for the final quadrat measurement.

Statistical analyses
Data were evaluated for statistical assumptions of normality and homogeneity of variance.
❖ www.esajournals.org Rank transformations were applied as necessary to achieve these assumptions, and the activityabundances of taxa and species were rank transformed to improve normality (Quinn and Keough 2002). For each sampling interval, trap catches were standardized to 14 trap days to account for disruption and the number of operational days [(trap catch/total number of days trap was operational) 9 14] (Spence et al. 1996, Gandhi et al. 2008. Taxonomic richness (number of invertebrate taxa per treatment), Pielou J evenness index (Pielou 1966), and Shannon diversity index (Peet 1974) were calculated using PC-ORD software (McCune and Mefford 2006).
Repeated measures analysis of variance (ANOVA) tests coupled with univariate analyses for each date and Tukey's pairwise comparisons for mean separation following a significant F test were used to compare grounddwelling invertebrate communities in canopy gap and understory vegetation disturbance treatments using SAS software (SAS 2014(SAS -2016. Response variables were total grounddwelling invertebrate activity-abundance, taxonomic richness, evenness, and diversity. Predictor variables for the models were canopy gaps (presence/absence) and understory vegetation (presence/absence) as fixed factors and sampling interval as a repeated factor. Each year was analyzed separately. If no pre-treatment differences were detected between experimental quadrats, total ground-dwelling invertebrate activity-abundance, taxonomic richness, evenness, and diversity were pooled over the summer in 2013 and used as covariates in the corresponding repeated measures ANCOVA tests for 2014 and 2015. Similar repeated measures ANOVA tests were used to assess the change in canopy and understory disturbance treatments over the three-year study, as well as compare the forest floor environment between these treatments. Response variables were percentage canopy openness; percentage cover of understory vegetation, leaf litter, bare ground, FWD, CWD, and rocks; height of understory vegetation; leaf litter depth; soil temperature; and soil moisture. Each year was analyzed separately. Logistic regression analyses with maximum likelihood estimates were conducted to compare canopy decline of girdled trees over time using the 1-5 rating scale.
Separate analyses were conducted to compare differences between tree species and between ring-and diffuse-porous trees.
Activity-abundances of ground-dwelling invertebrate taxa were analyzed to understand taxonspecific responses to the disturbance treatments. The activity-abundance of a taxon was analyzed if ≥20 individuals were collected during the year, and each year was analyzed separately. Activityabundances of common species (≥150 total individuals collected) in the families Formicidae, Carabidae, Nitidulidae, Geotrupidae, Scarabaeidae, Trogidae, and Silphidae also were analyzed. Responses were assessed only for common species to ensure these were not transient, but have established populations at the study site. Similar repeated measures ANOVA tests as described above were used to analyze the activity-abundances of ground-dwelling invertebrate taxa and common species using SAS software (SAS 2014(SAS -2016. Multivariate regression tree (MRT) analyses were used to investigate the relationship between ground-dwelling invertebrate community composition and the continuous ground-level environmental factors measured in the treatments. Multivariate regression tree is a constrained clustering method that is robust to complex ecological data that may contain missing values, nonlinear relationships between response and predictor variables, and interactions between predictor variables (Borcard et al. 2011). This technique repeatedly partitions the data into groups defined by thresholds in environmental variables (De'ath 2002, McCune andGrace 2002). Distance-based MRT (db-MRT) analyses were conducted separately for communities in 2014 and 2015 using the Bray-Curtis dissimilarity index. The size of each tree was selected by cross-validation using the ANOVA method to minimize the relative error and cross-validated relative error (De'ath 2002). Indicator species analyses (ISA) were conducted to identify invertebrate taxa that characterized each terminal group in the tree (De'ath 2002, Borcard et al. 2011). This technique is based on the relative abundance and relative frequency of taxa present in each quadrat (Dufrêne and Legendre 1997). Multivariate regression tree analyses were conducted using mvpart (De'ath 2014) and MVPARTwrap (Ouellete 2013) packages, and ❖ www.esajournals.org ISA were conducted using labdsv (Roberts 2016) in R version 3.3.1 (R Development Core Team 2017).

Forest floor environment
The forest floor environment was similar in all experimental quadrats in 2013 (Appendix S1: Table S2). After treatment implementation, soil temperature was consistently higher in the understory removed treatment (
Coleoptera was the most abundant insect order, comprising 33 families and totaling 31,112 individuals. Beetles in the families Carabidae, Geotrupidae, Scarabaeidae, Trogidae, Silphidae, and Nitidulidae were identified to species (Appendix S1: Table S1). Carabidae was the most diverse family with 5166 individuals comprising 30 genera and 59 species, of which Carabus goryi Dejean and Platynus angustatus Dejean were the most abundant. In the family Geotrupidae, species in the genus Geotrupes were abundant (Geotrupes semiopacus Jekel, Geotrupes splendidus [Fabricius], and Geotrupes balyi Jekel). Nine genera and 15 species were identified in the family Scarabaeidae, with Dialytellus striatulus (Schmidt) and Copris minutus Drury being the most common. Only one species, Trox variolatus Melsheimer, was collected in the family Trogidae. Nicrophorus orbicollis Say was the dominant species of Silphidae. Stelidota geminata (Say) and Glischrochilus sanguinolentus Olivier were common species in the family Nitidulidae.
Responses to either canopy or understory disturbance treatments were detected for 20 grounddwelling invertebrate taxa (Appendix S1: Table S3).
❖ www.esajournals.org Activity-abundances of Caseyidae, Paradoxosomatidae, Isopoda, Araneae, Isotomidae, Tomoceridae, Dicyrtomidae, Katiannidae, Gryllidae, Carabidae, Scarabaeidae, and Staphylinidae were higher in canopy gaps than in the closed canopy treatment. Curculionidae and Nitidulidae were less abundant in canopy gaps. Activityabundances of Lithobiomorpha, Hypogastruridae, Entomobryidae, Rhaphidophoridae, Geotrupidae, Scarabaeidae, and Staphylinidae were higher in understory removed than in the understory present treatment. Canopy gap and understory vegetation removal treatments interacted to affect the activity-abundances of 11 ground-dwelling invertebrate taxa (Appendix S1: Table S3). In 2014, Caseyidae and Julidae were most abundant in the canopy gap/understory removed treatment compared to all other treatment combinations. Parajulidae was least abundant and Spirobolidae (Table 1  was most abundant in the canopy gap/understory present and the closed canopy/understory removed treatments. Polydesmidae, and to a lesser extent Paradoxosomatidae, were least abundant in the closed canopy/understory present treatment. Activity-abundance of Hypogastruridae was lowest in the closed canopy/understory present treatment, and a similar trend was observed for Isotomidae. In 2015, snails were most abundant in the canopy gap/understory present treatment, while slugs were least abundant in the closed canopy/understory removed treatment. Two families of Coleoptera showed consistent patterns of activity-abundance across the two years following canopy and understory disturbance. Histeridae were least abundant in the closed canopy/understory present treatment compared to all other treatment combinations in 2014 (F 1,20 = 5.75, P = 0.023) and 2015 (F 1,20 = 13.7, P = 0.001). Scarabaeinae were most abundant in the canopy gap/understory removed treatment and least abundant in the closed canopy/understory removed treatment in 2014 (F 1,20 = 5.51, P = 0.028) and 2015 (F 1,20 = 4.79, P = 0.042).
The ground-dwelling invertebrate community was characterized by a two-branch db-MRT in 2014 that was selected 96 of 100 times by crossvalidation. The tree divided the community into two groups based on a threshold of 8.2% cover of FWD. Paradoxosomatidae (44.6%), Polydesmidae (20.2%), and Formicidae (15.0%) contributed most to the variance explained by the split, but model accounted for only 50% of the total variance, returning a high relative error of 50% and cross-validated relative error (AE SE) of 1.6 (AE 0.5), suggesting a poor fit overall. Percentage cover of FWD < 8.2% was observed for the majority (n = 26) of experimental quadrats, while fewer (n = 6; including the undisturbed control and canopy gap treatment) had ≥ 8.2% FWD cover.

DISCUSSION
Formation of canopy gaps had a greater effect on the structure and composition of grounddwelling invertebrate communities than did a pulse disturbance to the understory, which is consistent with our first prediction. This finding also is consistent with previous studies that found the formation of canopy gaps to be a Fig. 5. Distance-based multivariate regression tree (db-MRT) comparing ground-dwelling invertebrate communities in canopy and ground-level understory vegetation disturbance treatments in 2015 at Powdermill Nature Reserve, Westmoreland County, Pennsylvania. The three-branch db-MRT partitioned the invertebrate community based on soil moisture levels (%) and canopy openness (%), explaining 61.6% of the total variance. primary driver of invertebrate community structure (Richardson et al. 2010, Perry and Herms 2016a. Invertebrate activity-abundance and richness were consistently higher in canopy gaps. These patterns were primarily driven by increased activity-abundances of diverse functional guilds such as detritivores, fungivores, and predators that included families of Diplopoda, Collembola, and Coleoptera. Although these findings corroborate the importance of canopy gap formation for structuring invertebrate communities following disturbance, the positive effects of gaps on invertebrate abundance and richness reported here are inconsistent with other studies (Greenberg and Forrest 2003, Gandhi et al. 2008, Richardson et al. 2010. For example, canopy gaps experimentally created to emulate hurricane disturbance had lower diversity and biomass of ground-dwelling invertebrates such as predators and large detritivores compared to undisturbed forest (Richardson et al. 2010). Greenberg and Forrest (2003) observed lower total ground-dwelling invertebrate activityabundance and biomass in windthrow gaps created by Hurricane Opal than in nearby undisturbed forest, primarily due to lower activity-abundances of Carabidae, Araneae, Julida, Spirobolida, and Scolopendromorpha. Gandhi et al. (2008) documented lower activity-abundance and diversity of ground beetles in canopy gaps created by a windstorm than in undisturbed forest. In our study, families of Collembola were the most abundant taxa collected and were among the dominant taxa driving patterns in total ground-dwelling invertebrate responses. Greenberg and Forrest (2003) did not sample Collembola, and Richardson et al. (2010) pooled families for their analyses, which may have obscured family-specific responses and explain the divergent patterns in total ground-dwelling invertebrate activity-abundance. Alternatively, smaller size of the experimental quadrats compared to largerscale windstorms may have facilitated use of the disturbed patch by forest and open-habitat invertebrates, thereby increasing total activityabundance and richness.
Compared to undisturbed forest, light availability, height and cover of understory vegetation, and soil moisture levels were higher in gaps, and depth and cover of leaf litter were lower. These environmental changes on the forest floor induced by canopy gaps may have contributed to altered ground-dwelling invertebrate activity-abundance, richness, and community composition. The importance of canopy gaps as a driver of forest community structure and diversity is well documented (Runkle and Yetter 1987, Yamamoto 1992, McCarthy 2001, Schnitzer and Carson 2001, as is the suite of environmental changes that occur on the forest floor following gap formation (Gray et al. 2002, Ishizuka et al. 2002, Ritter et al. 2005, Perry and Herms 2016b. Canopy gaps were the primary factor structuring ground-dwelling invertebrate community composition in 2015, although the MRT model had low predictive power, but were not identified as a significant factor in 2014. Ring-porous tree species declined quickly following girdling in 2014, but these gaps were not large enough to drive significant changes in ground-dwelling invertebrate community composition, as indicated by the poor fit of the MRT model overall. In 2015, the model partitioned the quadrats according to the presence or absence of canopy gaps, which were larger after more diffuse-porous species declined and died. In the canopy gap treatment, families of epedaphic Collembola were numerically dominant and contributed most to the construction of the model, suggesting they responded strongly to environmental changes on the forest floor. Collembola are susceptible to desiccation and responsive to changes in soil moisture regimes (Verhoef 1977, Verhoef and van Selm 1983, Hopkin 1997, especially those species that are active on the litter and soil layers. Soil moisture was consistently higher in canopy gaps than in the closed canopy treatment, which may have influenced surfacedwelling Collembola populations. Henneron et al. (2017) also documented increased abundance and richness of epedaphic Collembola following canopy tree removal, but found that euedaphic Collembola (soil-dwelling species) decreased in abundance and richness. Regardless of their small size, Collembola assemblages are responsive to perturbations in forests and may be useful as an indicator taxon for environmental change.
The pulse of resource availability that occurs initially after gap formation may provide more opportunities for diverse invertebrate taxa. Tree ❖ www.esajournals.org death often increases soil moisture and nutrient levels (Minckler et al. 1973, Mladenoff 1987, Gray et al. 2002, Ishizuka et al. 2002. Increased soil temperatures and moisture levels create a conducive environment for decomposition, and increased mineralization and nitrification have been observed in forest openings (Likens et al. 1978, Runkle 1985. Detritivores and fungivores may have responded to this short-term resource pulse, and predators may have followed prey populations. Several epedaphic families of Collembola (Isotomidae, Tomoceridae, Dicyrtomidae, and Katiannidae), as well as common detritivores (Caseyidae, Paradoxosomatidae, Isopoda, and Scarabaeidae) and predators (Araneae, Carabidae, and Staphylinidae), were more abundant in canopy gap treatments. Soft-bodied arthropods such as Collembola are common prey for Araneae, Carabidae, and Staphylinidae, and their populations affect predator distributions (Chen and Wise 1999, Lawrence and Wise 2000, Birkhofer et al. 2010, Sereda et al. 2012. For example, the ground beetle species Chlaenius emarginatus, Cyclotrachelus sigillatus, and Pterostichus stygicus were more abundant in canopy gaps and may have recruited to these treatments due to abundant prey. All three species are predatory on soft-bodied invertebrates and are generalists in terms of their habitat requirements, as they are commonly collected in deciduous and mixed forest types as well as nearby fields (Larochelle and Larivi ere 2003). General habitat requirements may have primed these ground beetle species to take advantage of the abundant prey in canopy gap treatments, allowing them to remain in the disturbed patch. Patterns of movement of ground-dwelling invertebrates were mostly local in canopy gap treatments, as the majority of individuals dispersed only a short distance (<1-5 m) than those in undisturbed treatments (6-≥13 m; Perry 2016). This finding suggests that the majority of invertebrates remained within the disturbed patches (Perry 2016), perhaps due to the abundance of available resources in gaps.
This study investigated the impacts of disturbance intensity to the forest canopy and understory on ground-dwelling invertebrate communities. Therefore, treatments were implemented once and simulated a pulse rather than a press or chronic form of disturbance. Pulse disturbances that affect the canopy and understory such as windthrow are common in forest ecosystems (Oliver and Larson 1996, Bengtsson 2002, Frelich 2002, making this manipulative experiment a fairly good predictor for the impacts of natural disturbance events. However, because the disturbance treatments were only implemented once, temporal variability in canopy and understory conditions occurred during the course of the study. This variability may explain the poor fit of the MRT model characterizing ground-dwelling invertebrate community composition in 2014, as well as the dampened response of invertebrates to the understory removal treatment. The health of girdled trees declined over the summer in 2014, and the rate of tree decline was influenced by species-specific differences in vasculature structure (Wiant and Walker 1961), as ringporous species (oak, ash, and hickory) declined faster than diffuse-porous species (maple and poplar). Hence, gap size increased over time in all quadrats, but more rapidly in quadrats that had a greater number of girdled ring-porous trees. Gaps were larger in 2015 once the majority of trees had died, suggesting the magnitude of the treatment effect was greater than in 2014. This temporal change in canopy gap size explains the importance of gaps in structuring ground-dwelling invertebrate communities in 2015, but not 2014. The understory removal treatment displayed the opposite pattern with vegetation regrowth occurring over time, which decreased the effect size of the treatment. Therefore, the magnitude of understory vegetation removal decreased with plant regrowth, while the effects of gaps increased as tree canopies declined over time. Temporal changes in the disturbance treatments added an additional complexity to this study, and the responses of invertebrate taxa should be interpreted in light of these patterns.
Removal of understory vegetation affected ground-dwelling invertebrate communities, but these effects appear to be in response to vegetation recovery rather than its removal. Although differences in understory cover and height were greatest between undisturbed and removal treatments following implementation in 2014, invertebrate richness initially was unaffected. However, regrowth of the understory in removal treatments maintained higher invertebrate richness, even though understory cover became more similar between treatments. Detritivores, fungivores, herbivores, and predators increased following understory vegetation removal, similar to responses observed in canopy gaps, but these were largely different taxa. Neanuridae, Hypogastruridae, and Entomobryidae, all families of Collembola, increased following understory vegetation removal, even though movement of Collembola was lower (Perry 2016). Lithobiomorpha, a dominant predator on the forest floor, also was more abundant in the understory removal treatment. Woody plants were removed from the quadrats, but herbaceous plant residue remained on the forest floor to limit disturbance to the leaf litter layer. Additions of herbaceous organic material have been observed to stimulate microbial activity and soil nitrogen mineralization (Elliott et al. 2015), which may explain the increased richness of diverse invertebrate functional guilds in the understory removal treatment, as well as the rapid regrowth of vegetation.
The interaction between canopy and understory disturbances had less of an impact on grounddwelling invertebrates than either treatment alone. Responses to treatment interactions were detected only at the family and species level. Moreover, there was no evidence that treatment interactions influenced community structure more than canopy gaps alone. These results are inconsistent with our second prediction that invertebrate activity-abundance and diversity would decrease in response to combined canopy gap formation and ground-level understory vegetation removal.
Depending on canopy gap size, a dense layer of vegetation may develop in the understory following disturbance (Royo and Carson 2006), potentially mitigating the effects of abiotic changes induced by increased light on grounddwelling invertebrates (Shure and Phillips 1991). Although many taxa responded positively to the formation of canopy gaps, we were unable to detect an effect of the presence of ground-level vegetation due to the rapid growth response of the understory following removal in gap treatments. Therefore, our third prediction that the presence of understory vegetation in canopy gaps would increase invertebrate activity-abundance and diversity by buffering the variability in environmental changes on the forest floor could not be evaluated based on the current experimental design. Studies investigating responses of the understory herb layer following gap formation similar in size to this study found minimal differences in vegetation growth in canopy gaps compared to closed canopies Pickett 1987, 1988a, b), but these studies did not manipulate the understory layer. Rapid growth of the forest understory to levels of cover higher than pre-treatment conditions may have been in response to increased light and resource availability from the creation of canopy gaps, removal of vegetation, and addition of herbaceous residue.
Responses of ground-dwelling invertebrates to disturbance will depend to some degree on their ability to disperse within and between habitat patches, which likely influences patterns of community reorganization (Petraitis et al. 1989, Bengtsson 2002, Chase 2003. Dispersal potential of ground-dwelling arthropods was generally limited for most taxa , and movement was further reduced following combined canopy and ground-level understory disturbances utilized in this study (Perry 2016). Diplopoda were consistently responsive to canopy and understory disturbance, as well as their interaction, with the direction of change being largely family-specific. Activity-abundances of Caseyidae and Julidae increased in the most disturbed treatments with canopy gaps and the understory removed, whereas Polydesmidae and Paradoxosomatidae increased in all of the disturbance treatments. Millipedes generally displayed high dispersal potential ), but their movement, especially Paradoxosomatidae, was reduced in the most disturbed treatments with canopy gaps and understory vegetation removed (Perry 2016). Diplopods feed primarily on organic debris such as decaying plant material (Coleman et al. 2004, David 2009) that tends to have a patchy distribution on the forest floor. Therefore, they may be well adapted to exploit the pulse of resources in canopy gaps, affecting their patterns of movement.
A pulse disturbance event was emulated via a manipulative experiment, and responses of ground-dwelling invertebrates were comprehensively investigated before and after at the community, family, and species level, including those of key environmental indicator taxa. Decoupling of canopy and understory vegetation disturbances revealed canopy gap formation as an important short-term driver of ground-dwelling invertebrate community structure and composition, affecting the activity-abundance of diverse functional guilds, including predators, detritivores, and fungivores. Not only were these taxa more abundant in canopy gaps, they also tended to stay in the gaps, but with increased local movement (<1-5 m), as evidenced from a companion markcapture study conducted simultaneously in the same experimental quadrats (Perry 2016). Understanding how multiple components of natural and anthropogenic disturbances interact and impact forest communities can inform sustainable forest management practices, especially those that involve small-and large-scale canopy removal. As the frequency and intensity of strong storms are predicted to increase with climate change (Dale et al. 2000(Dale et al. , 2001, disturbance to forest canopies likely will become more common and widespread. These findings will improve understanding of how invertebrate communities respond to natural and anthropogenic disturbances, and how disturbance impacts biodiversity and resilience in forest ecosystems.

ACKNOWLEDGMENTS
We are very thankful to Amilcar Vargas-Loyo, Kara Taylor, Christian Bonilla-Sanchez, Diane Hartzler, Andrea Wade, Nicole Sharpe, Cali Granger, Paul Snyder, Sarah Rudawsky, (Ohio State University), and Ellen Thorne (PNR) for their extensive field and laboratory assistance. Christian Bonilla-Sanchez (Ohio State University), Sarah Pears, and Hanna Aronowitz (University of Vermont) assisted with treatment implementation. Cokie Lindsay facilitated travel and lodging at PNR. Robert Davidson, Robert Androw (Carnegie Museum of Natural History), and Joe Raczkowski (Ohio State University) provided species identifications and verifications. James Whitacre and Jake Slyder (PNR) assisted with GIS. Two anonymous reviewers provided helpful comments that improved the manuscript. Funding for this research was provided by The Ohio Agricultural Research and Development Center Graduate Research Enhancement Grant Program, Carnegie Museum of Natural History Graduate Student Research Grant, and state and federal funds appropriated to the Ohio Agricultural Research and Development Center and The Ohio State University.