Leaving windthrown stands unsalvaged as a management practice for facilitating late-successional carabid assemblages

The usual practice following a windthrow event is salvage logging of all damaged trees. This study was carried out in a pine stand affected by storms with varying disturbance severities in which no salvage logging was performed. Epigeic carabids (Coleoptera) were the test taxon. How does windthrow severity and disturbance legacies affect the abundance, richness and life traits of carabid assemblages? Two hypotheses were formulated: (1) the more severe the disturbance, the higher the abundance and species richness, and the lower the proportion of late-successional fauna in carabid assemblages, and (2) pine tree crowns lying on the ground support a higher proportion of late-successional carabid species, the effect of which should be more pronounced in more severely disturbed stands but will decrease as needles fall off over the next 3 years. To test the first hypothesis, 3 degrees of stand disturbance were distinguished. Both hypotheses were partially supported. Simultaneously, it was demonstrated that the importance of the leaf area index for carabids decreased over the 3 years, while the significance of the soil respiration rate and soil humidity increased. During the 3 years, the percentage of carabid forest species living under the fallen tree crowns remained at the same level. The findings indicate that there is a rationale for leaving fallen trees in place together with their crowns, while in tree stands where salvage logging is performed, whole crowns should be left cut-off from trunks to provide shelter for forest species.


Introduction
The frequency and intensity of windthrow disturbances (that is, the consequences of a wind storm that breaks or overturns trees over large areas) of forest stands in Europe and around the world have been increasing since the late twentieth century (Nilsson et al. 2004;Senf and Seidl 2021). Accordingly, research is now targeting not only the consequences of stand disturbance but also various ways to support stand recovery (Schönenberger 2002;Hotta et al. 2021;Wohlgemuth et al. 2017). The effects of a windthrow event depend on the force of the wind, species composition and height of the forest stand, soil type, soil depth, soil waterlogging, slope inclination and other factors (Dobbertin 2002;Bouchard et al. 2009). Windthrow disturbs stands over large areas, producing a patchwork of different degrees of disturbance severity, which entails different rates of stand regeneration (Skłodowski 2020). Severe windthrows revert succession in the stand to early developmental stages where pioneer species are dominant (Christie and Armesto 2003). In less disturbed stands with more surviving trees, established seedlings of tolerant species may facilitate forest stand recovery to late-successional stages (Girard et al. 2014).
Stand renewal most often starts after the removal of broken and upturned trees. Salvage logging facilitates the planting of tree saplings, which can reach higher density and height than emerging saplings from seeds present in the soil (Schönenberger 2002;Wohlgemuth et al. 2017). However, salvage logging causes the forest ecosystem to revert to an earlier successional stage and thus impedes recovery (Taeroe et al. 2019). Salvage logging also exerts a moderate negative effect on the ecosystem and its services (Leverkus Communicated by Claus Bässler. et al. 2020) and negatively affects the number of species and species richness of invertebrates (Thorn et al. 2018). Consequently, the structure of invertebrate assemblages undergoes significant changes consisting of a decreased proportion of forest specialists (late-successional fauna) and an increased proportion of the early-successional fauna of non-forest species (Otte 1989). By affecting the abundance and distribution patterns of resources on macro-and microhabitat scales in disturbed forest stands, a windthrow event impacts the abundance and species richness of insect species (Bouget and Duelli 2004).
Among the many invertebrate families, ground beetles (Coleoptera, Carabidae) are often used in ecological research owing to their high diversity and widespread distribution, their well-established ecology and easy trappability (Lӧvei and Sunderland 1996;Koivula 2011). The carabid response to various types of disturbance and stress (e.g. clear-cutting, forest fires, anthropogenic pollution) has often led to decreased abundance of late-successional fauna and increased abundance of early-successional fauna, producing an increase in species richness (Beaudry et al. 1997;Magura et al. 2001Magura et al. , 2015Magura et al. , 2017Koivula and Niemelä 2003;Paquin 2008;Nagy et al. 2016;Kędzior et al. 2018;Kosewska et al. 2018). The response of carabids to stand disturbance by windthrow has been reported in a number of studies (Duelli et al. 2002;Kašák et al. 2017;Skłodowski 2017a). Windthrow events in forest stands followed by salvage logging produce a marked reduction in the abundance of carabid forest species, whose recovery starts a few years after the stands re-establish. Conversely, in windthrown stands where all wind-felled trees were left in place (no salvage logging), the proportion of forest species in carabid assemblages remained quite high during the first three years after a disturbance but decreased only in the 4th year. This may have been associated with the possible survival of forest species under the fallen crowns of wind-felled pine trees (Skłodowski 2017a).
The crowns of windthrown pines retain their needles for several months, and the needles fall off gradually over 3 years (personal observation). Every crown with needles offers protection against excessive sunlight and over-drying of at least a dozen or so square metres of litter around the trunk. These are appropriate conditions for sheltering forest fauna (Bouget 2005). Similarly, logging residue and stumps in clear-cut sites support overwintering, egg laying and hiding of forest carabid species (Cobb et al. 2007). Stacks of branches also support the survival of late-successional carabid fauna even in the centre of a clear-cut during the first 2 years after clear-cutting (Skłodowski 2017b), whereas extensive slash removal impoverishes the species richness of ground-active beetles (Gunnarsson et al. 2004). Some authors suggest that slash harvesting produces a short-lived effect; as early as one year after stump harvesting, it had no effect on assemblage composition (Shevlin et al. 2017). At the same time, other studies have shown that salvage logging may have long-lasting effects on the composition and structure of carabid assemblages; even 5 years after cutting, the number of carabid species in stands was significantly higher in clear-cuts than in sites where slash was left on the ground (Nittérus et al. 2007). Slash removal caused a decline in forest species and an increase in generalist species (Nittérus et al. 2007).
The dispersal power of carabid species can vary from high to low. The former group comprises non-forest species that efficiently colonise disturbed habitats owing to their ability to fly, while the latter group consists of wingless (without functional wings) forest specialists that are only able to move along the ground and thus do not cover long distances. In a windthrown stand, members of this latter group, representing late-successional fauna, probably seek shelter in canopies of broken pine trees lying on the ground. As the occurrence of these species is supported by a greater degree of canopy closure (Magura et al. 2006;Taboda et al. 2008), the shade of pine crowns lying on the ground should facilitate the survival of forest species. It appears that the protective effect of fallen tree canopies on forest and latesuccessional species should be more pronounced in more disturbed stands (with all or nearly all trees broken) than in less disturbed stands (with half of the trees intact). However, as needles fall off in successive years, the protective effect presumably weakens. Hence, the following two hypotheses were formulated: (1) the more severe the disturbance, the higher the abundance and species richness, and the lower the proportion of late-successional fauna in carabid assemblages, and (2) pine tree crowns lying on the ground support a higher proportion of late-successional carabid species, the effect of which should be more pronounced in more severely disturbed stands but will decrease as needles fall off over the next 3 years. An additional objective was to identify the environmental variables that are the most conducive to the survival of late-successional carabid fauna under crowns of fallen pine trees in windthrown stands in consecutive years.

Study area
On the night of 11 to 12 August 2017, a storm disturbed 120,000 ha of forest stands managed by the Gdańsk Regional Forest District and Toruń Regional District in northern Poland. The storm reached a speed of 150 km/h and caused damage to forests in 60 forest districts. The volume of fallen trees was 9.8 million m 3 . It was the largest and most severe disturbance caused by a storm in Poland. In the area of the disturbed stands where the research was conducted, no salvage logging was 1 3 performed for the first 3 years. The disturbed pine stands were left uncleared for research purposes in the Lipusz State Forest Management Unit, forest divisions 154j, 155h, 155g, 155f, 155m, 157d and 157c (Fig. 1). Before the disturbance, Scotch pine (Pinus sylvestris L.) was the dominant tree species in the stands, with a marginal contribution from birch (Betula pendula Roth.). Before the windthrow event, the pine stands had 80% canopy closure. The height of the pine stand canopy was up to 25m. The stands grew on rusty-podzolic soil with a more litter layer. In the least disturbed stands, the most abundant species in the herbaceous layer were Vaccinium myrtillus L., Vaccinium vitis-idaea L. and Melampyrum pratense L. Deschampsia flexuosa (L.) Trin grew as isolated single plants. The moss layer consisted of Dicranum polysetum Sw. Ex Anon., Hylocomium splendens (Hedw.) Schimp. and Pleurozium schreberi (Willd. Ex Brid.) Mitt. In the moderately and severely disturbed stands, the occurrence of most of these species was reduced; only D. flexuosa and V. vitis-idaea increased in density.

Sampling design
To study the disturbance severity, three treatment types were chosen: severely disturbed (canopy cover of 0-20%; hereafter SD), moderately disturbed (canopy cover of 20-60%, hereafter MD) and barely disturbed, with just a few trees broken by windthrow (60-90%). The canopy cover percentage of surviving trees was calculated as the ratio of the number of surviving trees to the number of damaged trees. In the studied disturbed stands, broken and fallen trees laid on the ground, i.e. crowns with a larger part of the log (hereinafter referred to as crowns of fallen trees or pines). Additional treatments were established in SD and MD stands, where traps were placed under tree crowns lying on the ground. Thus, five treatments were distinguished: severely disturbed stands with (SD) and without (SCD) tree crowns lying on the ground (1 and 2), moderately disturbed stands with (MD) and without (MCD) tree crowns lying on the ground (3 & 4) and least disturbed (LD) stands (5). Each treatment was replicated three times for a total of 15 research plots (5 treatments × 3 replications) (Fig. 1).
Carabids were captured between 2018 and 2020 (1-3 years after the windthrow event) in pitfall traps (0.5-L glass jars with a plastic funnel, an upper diameter of 12 cm and a lower diameter of 2.0 cm, containing 200 ml of 70% ethylene glycol). The jars were placed in dug holes so that the upper rims of the funnels placed in them were flush with the litter layer. Four pitfall traps were set 10 m apart in each replicated treatment on 15 May, for a total of 60 traps (5 treatments × 3 replications × 4 traps). Traps were placed between fallen trees (Hypothesis 1) and under crowns of wind-felled pine trees (Hypothesis 2). The traps were subsequently replaced on 1 July, 15 August and 30 September. The procedure involved replacing the filled jar with a new, empty jar filled with glycol. The field work lasted 138 days each year.

Environmental variables
The variables determined as part of the study were the leaf area index (LAI-the total one-sided leaf area per unit of soil surface) in standing and fallen trees and the soil indices of temperature, humidity, respiration, pH value, and C and N content. LAI was measured directly at 10 cm above the soil with a LAI 2000 leaf area index metre (LI-COR Inc., Lincoln, NE, USA). In treatments with tree crowns lying on the ground, the LAI was measured at several locations within each of the crowns lying on the ground under which a trap had been placed. Soil pH and C and N contents were measured in soil samples collected from 20 randomly selected locations in each replicated treatment from a depth of 5 cm. The soil temperature and humidity were measured using a Pro-Check metre with a GS3 probe for measuring humidity (Decagon Devices, Inc. Pullman, USA). The soil CO 2 diffusion rate was measured directly in all treatments with an EGM-5 Portable CO 2 Gas Analyser (PP Fig.1 Map of study sites in Lipusz State Forest Management Unit; SD, SCD, MD, MCD-severely and moderately disturbed stands without and with tree crowns lying on the ground; LD: least disturbed forests; 1-3: replications of treatments. Numbers in italics are the numbers of forest divisions Systems, Amesbury, MA, USA). The soil pH was determined in water and in KCl solution (in the laboratory with a Thermo Orion 550A pH metre; Thermo Electron Corporation Environmental Instruments, Beverly, MA, USA). Carbon content was determined according to Tiurin's method (Mebious 1960) and nitrogen by Kjeldahl digestion (Bremner and Mulvaney 1982) in a certified laboratory (CentLab of the Institute of Environmental Protection-National Research Institute, Warsaw).

Carabid functional traits for analysis
Individual species of Carabidae were divided with regard to two life traits: preferred habitat (environmental groups) and diet (trophic groups), according to the current ecological knowledge about these species (Burakowski et al. 1973(Burakowski et al. , 1974Szyszko 1983;Hůrka 1996;Turin 2000). The habitat-based division comprised species inhabiting closed-canopy and mature forests (further referred to as forest species), generalist species (inhabiting forests and open areas) and species of open habitats. The trophic division comprised large zoophages (carnivorous species over 15 mm in body length), small zoophages (carnivorous species below 15 mm in body length) and hemizoophages (semi-carnivorous species). The percentages of individuals belonging to forest, generalist and open-habitat species and large zoophages, small zoophages and hemizoophages in the carabid assemblages were analysed. A high proportion of forest species and large zoophages, which can be considered successional fauna, suggests an advanced phase of successional development of the carabid assemblages and the habitat. Conversely, severely disturbed stands are colonized by hemizoophages, generalists and species of open areas, which can be regarded as early-successional fauna.

Data analysis
The catches in the 4 traps from each replication of a treatment were pooled. Species richness was standardized to the lowest number of specimens in the samples using Simberloff's rarefaction formula modified by Hurlbert (Krebs 1999 The first hypothesis, stipulating a reduced proportion of late-successional fauna with increasing windthrow severity, was tested using a generalized linear mixed model (GLMM) in R (ver. 3.5.0.; 'nlme', 'lsmeans' and 'car' packages) (R Core Team 2018). Treatments without canopies lying on the ground (plots SD1-SD3, MD1-MD3 and LD1-LD3) and time (consecutive years of study) were fixed effects, while replications of treatments constituted a random effect. A post hoc comparison of significant differences in GLMM was made using the Tukey significant difference test. The response variables (abundance and standardized species richness and carabid functional traits) were defined as following the Poisson distribution (with the log link function). Predicted and observed residuals were compared during the analysis to rule out any autocorrelation.
The second hypothesis, predicting a higher proportion of late-successional fauna in assemblages under crowns of pine trees lying on the ground (plots SD1-SD3, SCD1-SCD3, MD1-MD3 and MCD1-MCD3) than in open areas and a reduction in this proportion over time as needles fall off the canopies, was tested using a GLMM in R (ver. 3.5.0.; 'nlme', 'lsmeans' and 'car' packages) (R Core Team 2018). Treatments with (MD, SD) and without (MCD and SCD) tree crowns lying on the ground and time were fixed effects, and replications of the treatments were a random effect. The response variables (abundance and standardized species richness and carabid functional traits) were defined as following the Poisson distribution (with the log link function). A post hoc comparison of significant differences in GLMM was performed using the Tukey HSD significant difference test. Predicted and observed residuals were compared during the analysis to rule out any autocorrelation.
To enhance the testing of both hypotheses, the indicator value (IndVal) procedure (Dufrene and Legendre 1997) was used to analyse individual carabid assemblages to identify characteristic species. The IndVal method evaluates how strongly individual species are associated with pre-defined groups of sites in a set of data. The IndVal index attains its maximum value of 1.0 when all individuals of a given species are recorded in a single treatment (high specificity) and the species is found across all replications of that treatment (high fidelity). The statistical significance of the IndVal index was analysed with the Monte Carlo permutation procedure. The IndVal index was calculated in the R environment (ver. 3.5.0) ('labdsv' package, R Core Team 2018).
To attain the additional objective of identifying environmental factors that are the most conducive to the survival of carabid beetles under pine tree canopies lying on the ground, redundancy analysis (RDA) was performed using CANOCO 4.5 with pre-selected environmental variables using the ANOVA function with a permutation test (Ter Braak and Šmilauer 2003). The RDA was conducted after testing the length of the gradient (the length being lower than 3.0) in detrended canonical correspondence analysis (DCCA).

Dynamics of ground beetle assemblages in disturbed stands
The catches in 2018-2020 yielded 2381 individuals belonging to 65 species (see Appendix B for a species table). The highest individual counts were recorded for the following species: Amara lunicollis (708), Carabus violaceus (302), Carabus glabratus (225), Pseudoophonus rufipes (224), Pterostichus oblongopunctatus (205) and Calathus micropterus (150). The rarefied species number and abundance of carabids did not differ among assemblages inhabiting the different windthrow severity treatments (Table 1, Fig. 2a, b). Increasing stand disturbance severity was associated with an increasing proportion of generalist species and hemizoophages (Table 1, Fig. 2d, h). The proportion of forest species, large zoophages and small zoophages was significantly greater in the LD stands than in the MD or SD stands, while the proportion of open-habitat species was lower (Table 1, Fig. 2c, f, g, e). Between years 1 and 2 post-windthrow, the proportion of individuals of generalist species increased significantly, and the proportion of open-habitat species decreased significantly (particularly in the MD and SD stands) (Appendix C and D).

Ground beetles and pine crowns lying on the ground
The rarefied species number and abundance of carabids did not differ between assemblages living under crowns of fallen pines vs. those in areas without crowns of fallen pine trees (Table 1, Fig. 2a, b). The proportion of forest species, large zoophages and small zoophages was higher, and the proportion of generalist species and hemizoophages was smaller in the assemblages living under the crown of fallen pines than in the assemblages inhabiting treatments without crowns of fallen pine trees (  Fig. 2b, f, g, d, h). No significant differences in the proportions of forest species, small zoophages, large zoophages, generalist species, hemizoophages and open-habitat species were found for assemblages living under crowns of fallen pines in BD vs. MD and SD stands. Similarly, no significant changes in the proportion of any of these species groups were noted over the three years of the study (Appendix C and D).
The IndVal analysis for 2018 detected Agonum fuliginosum and Carabus hortensis as characteristic assemblages inhabiting pine canopies lying on the ground, Poecilus versicolor, A. lunicollis and P. rufipes as characteristic of MD and SD stands without tree crowns lying on the ground and three species (C. micropterus, Harpalus quadripunctatus and C. violaceus) as characteristic of LD stands ( Table 2). The IndVal analysis for 2019 indicated only A. fuliginosum as characteristic of the assemblages living under pine canopies lying on the ground and 4 species (Carabus arvensis, Carabus convexus, C. violaceus and C. micropterus) as characteristic of LD stands. The IndVal analysis for 2020 did not identify any species as characteristic.

Ground beetles and environmental variables
The first and second axes on the RDA diagram in the three consecutive years of the study explained 18.1%, 50.5%, 80.7% and 77.4%, 11.0% and 0.5%, respectively, of the variance in species data. The first and second axes on the RDA diagram during the three years of the study explained 100.0%, 78.4%,  Fig. 3). Forwards selection of explanatory environmental variables indicated the following as significant: in 2018, LAI (F = 2.87; p = 0.05, correlation with 1st species axis: r = − 1.0); in 2019, LAI (F = 7.77, p = 0.004, correlation as above: r = 0.73), soil temperature (F = 4.93, p = 0.004, as above: r = − 0.59) and C content in soil (F = 2.71, p = 0.036, as above: r = 0.18); and in 2020, soil respiration (F = 17.72, p = 0.004, as above: r = − 0.77) and soil humidity (F = 14.96, p = 0.002, as above: r = − 0.47). LAI, soil humidity, soil respiration and C content in the soil correlated with the assemblages from LD stands and those living under tree canopies lying on the ground in MCD and SCD stands.
In 2018, the 1st RDA axis separated a compact group on the diagram of carabid assemblages inhabiting LD stands and those living under pine tree canopies lying on the ground in MD and SD stands as distinct from the assemblages inhabiting MD and SD stands without tree crowns lying on the ground. In 2019, the RDA partially separated the assemblages inhabiting LD stands from those inhabiting tree crowns lying on the ground in MCD and SCD stands, increasing the distance between them. In 2020, the RDA partially separated these two groups of carabid assemblages, placing between them the assemblages inhabiting MD and SD stands without tree crowns lying on the ground.

Discussion
This study sought to answer the question "how does windthrow severity and disturbance legacies affect the abundance, richness and life traits of carabid assemblages?" Accordingly, two hypotheses were formulated: (1) the more severe the disturbance, the higher the abundance and species richness, and the lower the proportion of late-successional fauna in carabid assemblages, and (2) pine tree crowns lying on the ground support a higher proportion of late-successional carabid species, the effect of which should be more pronounced in more severely disturbed stands but will decrease as needles fall off over the next 3 years. The first hypothesis was partially confirmed, as with increasing severity of stand disturbance, only the proportion of latesuccessional fauna in carabid assemblages decreased, while the abundance and species richness did not increase. The second hypothesis was also partially confirmed, as a higher proportion of late-successional carabid species under fallen pine crowns was observed, but the expected gradual reduction in the proportion of this group with needle drop over 3 years was not confirmed.

Dynamics of ground beetle assemblages in disturbed stands
Severe disturbances in the form of clear-cutting and windthrow increase abundance and species richness (Niemelëa et al. 1993;Bouget and Duelli 2004;Bouget 2005;Huber and Baumgarten 2005;Gandhi et al. 2008). Accordingly, failure to identify a rise in these parameters in the carabid assemblages in this study in increasingly more windthrow-disturbed stands could be regarded as surprising. This is probably related to the fact that whole fallen trees were left in the disturbed stand. Occupying a considerable area of the ground, such trees possibly delayed the onset of colonisation of the disturbed area by early-successional carabid fauna for 3 years (Skłodowski 2017a). These convergent findings indicate the importance of leaving damaged trees with their crowns in place.
The response of carabid assemblage structures to the windthrow disturbance of tree stands was visible as early as the first year post-disturbance. Progressively decreasing proportions of forest species, small zoophages and large zoophages and increasing proportions of individuals of generalist species and hemizoophages were associated with increasing severity of windthrow disturbance. A rapid and strong response of carabids following forest disturbances has been observed in other disturbed stands (Bouget and Duelli 2004;Bouget 2005;Gandhi et al. 2008). As forest fauna benefit from greater canopy closure (Magura et al. 2006;Taboada et al. 2008), the carabid response was least marked in LD stands. In SD stands, the temperature of the soil, which is not protected by trees, is higher (Chen et al. 1999), producing conditions tolerated by generalists and hemizoophages (Kašák et al. 2017). This explains why such species dominated the carabid assemblages of MD and SD stands, which is evidence of profound alterations in assemblage structure resembling those seen on clear-cut sites (Schwerk and Szyszko 2007;Kosewska et al. 2018;Skłodowski 2014Skłodowski , 2017aMagura et al. 2015).
Contrary to the hypothesis assuming reduced possibilities for protection of late-successional fauna with the falling of needles off fallen tree crowns, during the 3 years of observation, the proportion of forest species in assemblages inhabiting MD and SD stands did not decrease. This should perhaps be ascribed to the species seeking shelter in tree crowns lying on the ground. The only changes during the three years were noted among early-successional species that were colonising the disturbed stands, with an increasing proportion of generalist species and a reduction in open-habitat species. Perhaps generalists were also using the tree crowns lying on the ground as a refuge and thus gained an advantage over open-habitat species.  (e), large zoophages (f), small zoophages (g) and hemizoophages (h) in the least disturbed stands (LD), moderately disturbed stands (MD) and severely disturbed stands (SD) with tree crowns on the ground (MCD and SCD); lower-case letters above the bars indicate differences between data collected (at p < 0.01; Tukey's test)

Effects of pine crowns lying on the ground
Differences in abundance and rarefied number of species in the assemblages living under and between crowns of fallen trees were neither substantial nor significant, which may point to good mobility of carabids moving between these two habitats. In contrast, the proportion of late-successional species in the carabid assemblages living under the fallen crowns of wind-broken pines was higher than that in the assemblages living between tree crowns, and the proportion of early-successional species followed the opposite pattern. Crowns of damaged trees offer shelter for late-successional fauna (Pearce et al. 2003;Bouget 2005;Šustek and Vido 2013). However, there were no differences in the proportions of late-successional species in assemblages living under tree crowns in MD vs. SD stands, which suggests that when carabids can seek shelter in tree crowns lying on the ground, they will not respond to the quantity (higher vs. lower) of damaged trees available or the degree of soil protection provided by intact trees. The structure of carabid assemblages was not affected by the presence of fewer vs. even twice as many tree crowns lying on the ground.
The proportion of forest species in the assemblages inhabiting fallen tree crowns did decrease in a non-significant manner at one year after the windthrow event, but contrary to the hypothesis, it was not further reduced in the following years, remaining as high as the proportion of forest species in the assemblages inhabiting LD stands. This suggested that the living conditions for forest fauna under crowns of fallen trees were as good as those in LD stands even after the needles had fallen off. The more needles were shed, the thicker the layer of litter was under the crowns lying on the ground. Small tufts of moss could be seen under many of the fallen crowns, suggesting increased humidity of the habitat. Both the litter layer and mosses are excellent living habitats for forest carabids (Koivula et al. 1999;Magura et al. 2000), which are most often mesophilous and sometimes hygrophilous. A 2-5-cm-thick layer of shed needles under tree crowns lying on the ground may thus have supported the sheltering of individuals of forest species preferring the litter layer (Guillemain et al. 1997;Koivula et al. 1999;Koivula 2001;Magura 2002;Magura et al. 2001;Vican et al. 2018). It appears that these conditions were more beneficial to late-successional Fig. 3 Redundancy analysis of carabid assemblages; data from 2018 (a), 2019 (b), 2020 (c) for severely (S), severely + tree crowns on the ground (SC), moderately (M), moderately + tree crowns on the ground (MC) and least (L) disturbed stands. Only significant habitat variables are shown: LAI: leaf area index; temperature: soil temperature; C: soil content of carbon; CO 2 : soil respiration and soil humidity-measured directly in soil. The abbreviations of the species names are shown in the table in Appendix B. RDA model calculated without data transformation, with interspecies correlation scaling, species scores divided by SD and centring by species ▸ carabid fauna than the conditions offered by timber residues in clear-cuts, which provide a "natural" refuge (Nittérus et al. 2007;Skłodowski 2017a;Koivula et al. 2019).
In the first year post-windthrow, the species classified as characteristic of areas between tree crowns lying on the ground were the early-successional species P. versicolor, A. lunicollis and P. rufipes, which prefer dry and sunny habitats (Hůrka 1996;Turin 2000). Characteristic species in assemblages living under fallen tree crowns were the mesophilous and moderate hygrophilous species A. fuliginosum and C.
hortensis, which indicates that there was persistent humidity. This suggestion is all the more probable, as the small zoophage A. fuliginosum occurs in swamp forests (Skłodowski 2006), and the large zoophage C. hortensis, whose movement was traced in a forest-fallow ecotone, avoided moving out of the forest and into the much drier open area of the fallow field (Skłodowski 1995).
In LD stands, characteristic species were forest specialists: the small zoophage C. micropterus, large zoophage C. violaceus and hemizoophage H. quadripunctatus. C. micropterus prefers needle litter (Niemelä et al. 1988(Niemelä et al. , 1994(Niemelä et al. , 1996, which was the type of litter that survived in this treatment. In the managed forests of Białowieża Primaeval Forest, this species showed a preference for dense and compact 15-to 40-year-old coniferous stands. The large forest zoophage C. violaceus prefers forest litter (Magura et al. 2000) as well as moderately humid, coniferous or mixed forests and occurs at stand edges (Turin et al. 2003). In the forests of Białowieża, it showed a preference for coniferous stands in the primaeval forest over those found in the managed forests (Skłodowski 2006). Conversely, H. quadripunctatus is found in many managed forests (Szyszko 1983) and even in primaeval forests (Skłodowski 2006), but its abundance is low. It also occurs at forest edges (Hůrka 1996). As only some trees were broken in LD stands, this habitat might have resembled a forest edge to H. quadripunctatus.
In the second year post-windthrow, A. fuliginosum was the only species determined to be characteristic of the assemblages living under tree crowns lying on the ground. In that year, no species were identified as characteristic of the areas between tree crowns on the ground. The species characteristic of LD stands were forest species, including two known from the previous year (C. violaceus and C. micropterus) and two large zoophages (C. arvensis and C. convexus) indicated as characteristic for the first time. The xerothermal C. arvensis occurs in Central and Eastern Europe in thinned coniferous  (Turin et al. 2003), but it does not avoid leaving the forest and penetrating the open area of fallow land or clear-cuts (Skłodowski 1995(Skłodowski , 2008, with a preference for pine stands (Szyszko 1983). The mesophilous C. convexus has been recorded in Central Europe in various types of stands and at their edges (Turin et al. 2003). Accordingly, the presence of these two species as characteristic of LD stands suggests a thinning of these stands. In the third post-windthrow year, no species were identified as characteristic of the study treatments, which may point to increasing similarity in species composition among the carabid assemblages.

Ground beetles and environmental variables
In the first year post-windthrow, LAI was the only significant factor differentiating the carabid assemblages, which indicated the large importance of needles on pine crowns lying on the ground. Similarly, a high degree of canopy closure influences carabid assemblages (Purchart et al. 2013).
In the second year, LAI remained a significant factor, but some other factors were also significant, namely, high soil temperature (in areas between tree crowns on the ground) and soil C content (for carabids living in LD stands and under tree crowns). Higher C content in the organic soil layer is conducive to forest carabid assemblages (Schwerk et al. 2020) because it is associated with greater species diversity of litter and soil micro-and macrofauna, which results in better trophic conditions for forest species of ground beetles associated with forest litter.
In the third year of observation, LAI was no longer a significant factor, as most needles had fallen off the tree crowns. Significant factors for carabids in that year comprised a high soil respiration rate and high soil humidity under tree crowns lying on the ground. Higher soil humidity and respiration rate suggest efficient soil processes, good microbiological activity and, consequently, mineralization of organic matter (Vanhala et al. 2005). A higher soil respiration rate is seen in old stands with a thicker organic layer (Vanhala et al. 2005). This was why a layer of litter composed of shed needles exerted a significant influence on the carabid assemblages (Purchart et al. 2013).

Management conclusion
This study showed that late-successional carabid species can survive following a windthrow disturbance under the crowns of fallen pine trees lying on the ground, at least for the first 3 years since a severe windthrow event. Leaving tree crowns of fallen trees in place on the ground produces shelters for forest-associated species. At such sites, appropriate humidity and soil efficiency persist even after the needles have fallen off, as evidenced by a higher soil respiration rate and C content. Such sites represent valuable refuges for other invertebrates in addition to carabids. Accordingly, some broken-off crowns should be left in place during clearing of a windthrow-disturbed forest stand. As the MD and SD stands supported carabid assemblages with identical structures, there may be a rationale for leaving fallen trees together with their crowns in place, while in tree stands where salvage logging is performed, whole crowns should be left cut-off from trunks.

Appendix A
Average species number of carabids in the least disturbed stands (LD), moderately disturbed stands (MD) and severely disturbed stands (SD) with tree crowns of fallen trees (MCD and SCD) and average rarefied species number of carabids in the least disturbed stands (LDrar), moderately disturbed stands (MDrar) and severely disturbed stands (SDrar) with tree crowns of fallen trees (MCDrar and SCDrar).

Appendix B
Carabid check-list and species distribution in the least disturbed stands (LD), moderately disturbed stands (MD) and severely disturbed stands (SD) with tree crowns on the ground (MCD and SCD); species names according to Löbl and Smetana (2003). In the "RDA" column, the abbreviations of the species shown in the RDA analysis are entered (Fig. 3). zoophages and hemizoophages in carabid assemblages in SCD, MCD stands with tree crowns and without tree crowns SD and MD (treatment) over 3 years of observation (year).

Appendix D
Rarefied number of species (a), abundance (b) and proportion of individuals representing: forest species (c), generalist species (d), open-habitat species (e), large zoophages (f), small zoophages (g) and hemizoophages (h) over 3 years of observations in the least disturbed stands (LD), moderately disturbed stands (MD) and severely disturbed stands (SD) with tree crowns on the ground (MCD and SCD); lower-case letters above the bars indicate differences between data collected (at p < 0.01; Tukey's test).