Effects of plantation intensity on longhorn and carabid beetles in conifer plantations mixed with broadleaved trees in northern Japan

ABSTRACT Reconciling timber production with biodiversity conservation is essential. Increasing the mixture of broadleaved (BL) trees into conifer plantations increases the abundance of species that prefer natural BL forests but can reduce conifer yields. Therefore, modeling of the relationships of the abundances of various taxa with conifer and BL trees is necessary for effective biodiversity conservation. Longhorn and carabid beetles are useful ecological indicators; however, their responses to the amount (e.g. basal area and coverage) of conifer and BL trees within stands remain unknown. We surveyed the abundances of longhorn and carabid beetles in plantations of Todo fir Abies sachalinensis and Sakhalin spruce Picea glehnii mixed with various amounts of BL trees and in natural BL forests. We analyzed the response of each taxon to the basal area of conifer trees (CBA) within stands. Unexpectedly, for longhorn beetles, no effect of CBA was detected for species using BL trees as larval host plants and the effect was positive for generalist species. For carabid beetles, the total abundance of forest-dependent species decreased almost linearly with increasing CBA. The species-level mean response was decreasing abundance with increasing CBA, which occurred more rapidly in the lower range of CBA. These results suggest large negative impacts of conifer trees, even in small amounts, for many carabid species. To reconcile timber production with carabid beetle conservation, it is essential to maintain remnant natural BL forests and revert unsuccessful conifer plantations containing abundant BL trees into BL forests.


Introduction
Natural forests are being lost, while planted forests or plantations are expanding globally (FAO 2020). Plantations generally support lower abundance and species richness of forestdependent species than old natural forests (Newbold et al. 2015;Hua et al. 2022;Wang et al. 2022) due to their lower diversity in terms of forest composition and structure (Carnus et al. 2006;Brockerhoff et al. 2008). Planted forests account for 36% of the total forest area in East Asia (FAO 2020), and the proportion is higher in lowland areas with high productivity and potential for biodiversity conservation (Yamaura et al. 2020). Thus, plantation expansion is a cause of biodiversity loss. However, the biodiversity of plantations can be enhanced through diversification of stand structure and composition (Hartley 2002;Demarais et al. 2017). Hence, both protecting natural forests and managing plantations have been acknowledged as important measures for regional biodiversity conservation (Yamaura et al. 2012;Demarais et al. 2017;McFadden and Dirzo 2018).
In conifer plantations, which are common in boreal and temperate zones (FAO 2006), adding broadleaved (BL) trees provides useful structures and resources for numerous species (Ohsawa 2007;Kebrle et al. 2021). Furthermore, mixing of BL trees into conifer plantations strongly influences the light, water, and soil environments, leading to differences in understory vegetation (Barbier et al. 2008). However, increasing the density of BL trees and decreasing that of conifers in plantations can reduce profit from timber production of planted conifers (Yoshida et al. 2005). Thus, trade-offs generally exist between the function of biodiversity conservation and the management intensity (i.e. plantation intensity, e.g. the density of planted trees; Betts et al. 2021). Modeling of the relationship between biodiversity and plantation intensity is necessary to identify the optimal solution that balances timber production with biodiversity conservation (Yamaura et al. 2012;Betts et al. 2021).
Two main strategies are used to ensure a mix of BL trees in conifer plantations: maintaining certain amounts (e.g. basal area [BA] and coverage) of BL trees in each conifer plantation (the land-sharing strategy) and reserving some stands with abundant BL trees around plantations for biodiversity conservation (the land-sparing strategy; Betts et al. 2021;Felton et al. 2022). Maintaining some BL trees in stands can aid the conservation of animal species that increase in abundance when BL trees are present even in small amounts but is unsuitable for species that require abundant BL trees (Yamaura et al. 2012;Betts et al. 2021).
Based on recent studies of birds, responses to the amount of BL trees likely differ among regions. In Finland, it has been shown that stands dominated by birch (Betula spp.) supported more abundant birds than did mixed stands of Norway spruce Picea abies and birch (Felton et al. 2021). Contrastingly, retaining some BL trees increased the total abundance of forest birds over the harvest rotation cycle in Todo fir Abies sachalinensis plantations in Hokkaido, northern Japan (Yamaura et al. 2023). In another area of Hokkaido (our study area), a moderate response (i.e. a linear increase with the BA of BL trees) was reported (Yoshii et al. 2015). However, not only region but also the dispersal abilities or dependency on BL trees of the conservation target species may affect biological responses to the amount of BL trees. Thus, it is necessary to explore the effects of plantation intensity on a range of taxa.
Numerous insect species depend on BL trees. For flying longhorn beetles, which have high mobility, larvae use dead wood and adults use various plant parts as food resources, but each species prefers different tree species (Ohsawa 2004;Maleque et al. 2009). Many ground-dwelling carabid beetles, which have low mobility, are supported by BL litter, which functions as their food source and shelter (Koivula et al. 1999;Magura et al. 2003;Skłodowski et al. 2018). Carabid beetles are also strongly influenced by canopy trees through effects on light, soil, and understory vegetation (Rainio and Niemelä 2003;Yu et al. 2010). Thus, these beetles are useful for examining the response to the amount of BL or conifer trees. Previous studies have demonstrated the effectiveness of maintaining BL trees within plantations or BL forests around plantations for conservation of these beetles (longhorn beetles: Maeto et al. 2002;Makino et al. 2007;Ohsawa 2007;carabid beetles: Yu et al. 2006carabid beetles: Yu et al. , 2010carabid beetles: Yu et al. , 2014Skłodowski et al. 2018). However, to our knowledge, no studies have yet explored the effects of quantitative differences in BL or conifer trees among stands in a continuous manner.
In Japan, plantations of conifer species (Cupressaceae [cypress family] and Pinaceae [pine family]) have largely replaced natural forests (Ueno et al. 2022), accounting for 41% of the nation's total forest area (Yamaura et al. 2012;Forestry Agency 2017). In pine family plantations, invasion by BL trees is more prevalent than in cypress family plantations (Yamaura et al. 2019), and the abundance and species richness of a range of taxa does not significantly differ from those in natural forests (Kawamura et al. 2021). However, few studies have examined the effects of BL tree amounts in plantation stands (but see Ohsawa 2007;Yoshii et al. 2015). Even fewer studies on insects have evaluated the habitat function of plantations of Todo fir and Sakhalin spruce Picea glehnii, which are common in Hokkaido, northern Japan (Kawamura et al. 2021).
In this study, we surveyed the abundance of longhorn beetles (BL-dependent and generalist species) and carabid beetles (forest-dependent species) in Todo fir and Sakhalin spruce plantations with different amounts of BL trees and in natural BL forests. We examined the responses of each group to the BA of conifers within stands, an important indicator used in forestry.

Study area and sites
Our study was conducted in plantations of Todo fir and Sakhalin spruce and natural BL forests in the Chitose National Forest in central Hokkaido (42°44′-42°50′N, 141°22′-141°37′E; Fig. S1). In Hokkaido, the levels of Todo fir and Picea (Sakhalin spruce and Yezo spruce Picea jezoensis) as proportions of the total plantation area were 52.2% and 11.5%, respectively (Department of Fisheries and Forestry of Hokkaido 2022). The study area is a mosaic landscape of natural forests and plantations, and the proportion of Sakhalin spruce plantations is higher than in other regions of Hokkaido (Fig. S1). Natural forests are composed mainly of deciduous BL trees, Japanese oak Quercus crispula, painted maple Acer pictum, monarch birch Betula maximowicziana, Japanese maple Acer palmatum, and Korean whitebeam Aria alnifolia (Yoshii et al. 2015). The flat terrain of our study area allowed us to select stands with similar conditions beyond the BA of conifers.
We selected seven Todo fir plantations and seven Sakhalin spruce plantations with different levels of mixed BL trees and four natural forests as survey sites (Fig. S2). We focused on homogenous stands of greater than 1 ha, and surveyed a 25 m × 25 m plot at each site. To reduce stand age effects, mature plantations were selected (stand age, Todo fir: 43-66 years, Sakhalin spruce: 35-48 years). The stand age of natural forests was approximately 50 years for two sites and more than 100 years for the other two sites. Among these natural forests, three sites were dominated by BL trees (100%, 100%, and 95% based on BA), and one site had 23% conifers (mainly Todo fir) by BA. We aimed to make each site spatially independent by setting a minimum distance of 300 m between sites. The Todo fir plantations were surveyed in 2020, Sakhalin spruce plantations in 2021, and natural forests in both years.

Beetle collection and grouping
We collected longhorn beetles using Malaise traps (Urabe-Kagaku, Sapporo, Japan: 1.5 m in height, 2 m in length, and 1.5 m in depth; Fig. S3a). We deployed one trap in the center of each plot, with a bottle for capturing insects at a height of 1.8 m. The traps contained ethanol diluted to 50% as a preservative. We collected insects every 2 to 3 weeks. Surveys were conducted from June 25 to 12 September 2020, and from July 8 to 22 August 2021.
For carabid beetle collection, we used pitfall traps ( Fig.  S3b). We set 15 traps (plastic cups, 74 mm diameter and 86 mm depth) along each of two lines in each plot (30 total traps per plot). The traps along each line were located 2 m apart, and the distance between lines was 10 m. The traps contained 60 mL propylene glycol diluted to 50% as a preservative. We collected insects 1 to 2 weeks after deployment of the traps. For sites where fewer than 10 traps remained operational due to animal attacks, we surveyed again. We surveyed twice in each plot (2020, first: from June 25 to July 31, second: from August 13 to September 12; 2021, first: from July 8 to July 16, second: from August 15 to August 22).
We pooled samples of both longhorn and carabid beetles by site and collection date. Samples were sorted and identified to the species level according to the literature (Table S1,  S2). For longhorn beetles, we categorized species into BL-, conifer-, and grass-dependent species, or as generalists based on their larval host plants, identified from the literature (Table S1). For carabid beetles, we categorized species into two groups, namely those preferring forests as habitat (forest-dependent species) and other species according to previous studies and a database (Table S2). We focused on BLdependent species and generalists among longhorn beetles and forest-dependent species among carabid beetles as conservation targets.

Vegetation and soil moisture survey
To determine the basal areas of conifers (CBA) and mixed BL trees (BLBA) in each plot, tree surveys were conducted from 12 September to 22 September 2020, and on 30 July 2021. We identified all trees with diameter at breast height > 5 cm to the species level, and recorded the diameter at breast height and height of each. Then, we calculated the CBA and BLBA (Table  S3). We also surveyed the understory vegetation volume and soil moisture content, which can affect the abundance of carabid beetles (Table S4). The understory vegetation volume was surveyed on the same day as the tree surveys. Three 2 m × 2 m quadrats were randomly placed in each plot, and the cover and maximum height of each plant species were recorded. The product of the cover and maximum height in each quadrat was calculated as the volume for each plant species, and these values were summed for all species to calculate understory vegetation volume. We used the mean value among three quadrats in each plot. Soil moisture content was measured using a time domain reflectometer (HydroSense, Campbell Scientific, Inc.), with three random replicate measurements at a depth of 10 cm below the ground surface in each plot, and the mean value of the three points was used for each site. These measurements were conducted at all sites on the same day after three or more days of clear weather (30 July 2020, and 22 August 2021).

Statistical analysis
To reveal the relationship between plantation intensity and beetle abundance, we employed generalized linear mixed models (GLMMs; family = Poisson, link function = log) in which the response variable was the abundance of each group (i.e. BL-dependent and generalist species of longhorn beetles and forest-dependent species of carabid beetles) at each site on each collection date; the explanatory variable was a value related to CBA (hereafter, group-level analysis). We constructed three types of models using the CBA at site i (CBA i ) and its maximum value (CBA_max) as explanatory variables: a model using the simple term CBA i (linear model), a model using log(CBA i + 1) (log model), and a model using log(−CBA i + CBA_max + 1) (log minus model). For beetle abundance, use of the log and log minus models enabled us to characterize a linear increase or decrease with CBA as well as a rapid increase or decrease within a narrow range of CBA (Fig. S7, S8). Following a previous study on birds in the same area (Yoshii et al. 2015), natural BL forests were treated as stands with low plantation intensity. Thus, data from natural forests were analyzed along with data from plantations in each model.
We used random intercepts in each model as follows: site ID for considering spatial differences in beetle abundance, which can be driven by factors other than CBA (random site effect), and season ID for considering seasonal changes in beetle abundance (random season effect). Assuming that the captured beetle abundance was directly proportional to sampling effort, we used the logarithmically transformed number of sampling days for longhorn beetles and the logarithmically transformed product of the number of valid traps and sampling days for carabid beetles as offset terms. As the effects of CBA can differ with the planted tree species, we analyzed the data from Todo fir and Sakhalin spruce plantations separately, and removed data from one natural forest site where Todo fir was present from the analysis of Sakhalin spruce plantations. Thus, data from 11 stands (7 plantations and 4 natural forests, surveyed in 2020) were used in the analysis of Todo fir plantations, and data from 10 stands (7 plantations and 3 natural forests, surveyed in 2021) were used in analysis of Sakhalin spruce plantations. We constructed a null model containing only the intercept in addition to the three models using CBA described above, and selected the model with the lowest Akaike information criterion (AIC) value as the best model (Yamaura et al. 2009).
As sampling effort differed among surveys, species richness could not be assessed through the same approach as abundance (i.e. we could not assume direct proportionality of sampling effort and species richness; Yamanaka et al. 2019). Thus, to reveal common patterns across many species, we conducted multi-species analysis using models in which the response variable was the abundance of each species. In this modeling process, the species-level mean response was estimated to be a fixed effect, and the differences in responses among species were the random intercept or slope effects (i.e. random species effects; Lehikoinen et al. 2019;Senzaki et al. 2020). As for group-level analysis, we constructed and compared four models (i.e. three models using CBA values as explanatory variables and a null model), in which data from the plantations that differed in tree species were assessed separately and the random site effect was considered. We considered a random seasonal effect for each species, as seasonal changes in abundance depend on species (i.e. the interaction term between season ID and species ID was added as a random intercept). For longhorn beetles, many species occurred only at specific sites. We analyzed only species that were captured at three or more sites to ensure model robustness. All analyses were conducted with R ver. 4.1.0 (R core team 2021) using the glmer function in the "lme4" package (Bates et al. 2015).

Results
For longhorn beetles, 117 individuals of 19 species were captured in 2020, and 165 individuals of 19 species were collected in 2021 (Table S1). In both years, the dominant species was Prionus insularis, a generalist species (total abundance: 39 in 2020 and 66 in 2021). For BL-dependent species, 18 individuals of eight species were used to analyze Todo fir plantations, and 31 individuals of six species were assessed from Sakhalin spruce plantations. Among them, 9 individuals of two species and 25 individuals of two species were used in multi-species analyses of the fir and spruce plantations, respectively. For generalists, 89 individuals of 9 species were used to analyze fir plantations and 129 individuals of 10 species were used to analyze spruce plantations. Among them, 76 individuals of four species and 125 individuals of six species were used in the multi-species analyses of fir and spruce plantations, respectively.
For carabid beetles, 4411 individuals of 22 species were captured in 2020, and 5079 individuals of 24 species were captured in 2021 (Table S2). In both years, the dominant species was Pterostichus thunbergi (total abundance: 2663 in 2020 and 3386 in 2021). For forest-dependent species, 4368 individuals of 15 species were used to analyze the fir plantations and 4984 individuals of 18 species from spruce plantations were assessed.

Responses of longhorn beetles to the basal area of conifer trees within stands
For BL-dependent species, the log model was selected as the best model in group-level analysis of the spruce plantations but its performance did not greatly differ from the null model (ΔAIC < 2; Table S5a). In group-level analyses of the fir plantations and multi-species analyses of plantations of both tree species, the null model was selected, and no effect of CBA was detected (Table S5b-d).
By contrast, for generalists, the linear model was selected as the best model in group-level analyses of spruce plantations, showing an unexpected response of abundance increasing with increasing CBA (Figure 1a, Table S6a). For fir plantations, the null model was selected, and no effect of CBA was detected (Figure 1b, Table S6b). Multi-species analyses of plantations of both tree species suggested positive effects of CBA. For spruce plantations, the log minus model was selected, indicating that beetle abundance increased with increasing CBA, and this increase was more rapid at higher values of CBA for all species (Figure 1c, Tables S6c and S7a). For fir plantations, the linear model was selected, and the coefficient of CBA was small but positive ( Figure 1d, Table S6d). Estimates of random species effects indicated that the response differed among species (Table S7b). The abundance of Prionus insularis (species ID: 20), the dominant species, decreased with increasing CBA. By contrast, the abundances of Leptura ochraceofasciata (species ID: 17) and Stictoleptura succedanea (species ID: 21) increased with CBA, and this increase became more rapid at higher values of CBA.

Responses of carabid beetles to the basal area of conifer trees within stands
In group-level analyses of plantations of both tree species, the linear model was selected as the best model, although its performance did not differ greatly from the null model (ΔAIC < 2; Table 1a, b). The best models showed that abundance decreased almost linearly with increasing CBA (Figure 1e, f). For multi-species analyses of plantations of both tree species, the log model was selected, indicating that abundance decreased with increasing CBA, and decreased more rapidly at lower CBA values (Table 1c, d, Figure 1g, h). This pattern of rapid decrease in the lower range of CBA values was more apparent for spruce plantations than fir plantations (Figure 1g). Species-level coefficients, calculated by adding random species effects to the fixed effect, were negative for many species but differed greatly among species ( Table 2), suggesting that not all species were sensitive to CBA and that the degree of negative impact from CBA differed among species. For Pterostichus thunbergi (species ID: 13), the coefficient of log(CBA i +1) had small negative values, indicating that abundance decreased gently with increasing CBA (Table 2). By contrast, larger negative coefficient values were estimated for Carabus albrechti albrechti (species ID: 1) in analyses of plantations of both tree species and for Carabus granulatus yezoensis (species ID: 5) and Synuchus melantho (species ID: 16) in the analysis of spruce plantations. Thus, increasing CBA, even in its lower value range, led to decreases in the abundances of these species (Table 2).

Figure 1.
Estimated relationship between the abundance of each insect group and basal area of conifers: (a -d) generalist longhorn beetles, (e -h) forestdependent species of carabid beetles, showing group-level analyses (response of total abundance) for Sakhalin spruce plantations (a, e) and Todo fir plantations (b, f), and multi-species analyses (species-level mean response of abundance) for spruce plantations (c, g) and fir plantations (d, h). Solid lines represent the expected value obtained from the best model. Translucent dots in (a, b, e, f) represent observed abundances at each site on each collection date.

Responses of longhorn beetles to the basal area of conifers within stands
For broadleaved tree-dependent species, the response to conifer BA was unclear. Previous studies have shown that the distribution of longhorn beetles is strongly affected by distributions of trees or forests at the landscape scale (Økland et al. 1996;Gibb et al. 2006). Thus, the effects of conifer amount within stands may have been masked by landscape-scale forest composition effects, which were not considered in this study. In our data, the abundance of broadleaved tree-dependent species was low in natural forests, potentially due to the negative impact of plantation expansion across the landscape (Bouget and Parmain 2016). Future studies on the responses of longhorn beetles to plantation intensity should consider such factors at multiple scales simultaneously.
For Sakhalin spruce plantations, generalists were abundant in stands with high conifer BA. In such stands, total tree BA was also high. Generalists can use conifers as egg-laying sites, and include species that have been frequently recorded from the wood of conifers (e.g. Prionus insularis and Stictoleptura succedanea ;Sasaki 2004;Ohbayashi and Niisato 2007). Therefore, this group is likely to be dominant in the plantation landscape. However, Ohsawa (2007) revealed that the species richness and abundance of longhorn beetles were higher near isolated, old broadleaved trees in plantations. Although we did not detect a response of broadleaved tree-dependent species to the amount of conifers within stands, their responses to factors at the individual tree scale, such as species, size, and deadwood amounts of broadleaved trees, or other stand-scale factors (e.g. light intensity or humidity), require detailed examination. Table 1. Results of GLMMs for the abundance of forest-dependent carabid beetles: group-level analyses of Sakhalin spruce plantations (a) and Todo fir plantations (b), and multi-species analyses of spruce plantations (c) and fir plantations (d). The values of the intercept and coefficients are presented as mean ± standard error (SE). Bold indicates the best model.  Table 2. Parameter estimates for each forest-dependent species of carabid beetles, derived from the best model for multi-species analysis of (a) Sakhalin spruce plantations, and (b) Todo fir plantations. The estimated value and standard deviation (SD) of the random intercept and slope for each species are shown. Low values for the intercept indicate that the mean abundance of the species was low. Species-level β indicates the slope of each species, calculated by adding the random species effect to the fixed effect. Bold: β < 0 (abundance decreased with increasing CBA, more rapidly in the lower range of CBA), underlined: 0 < β < 1 (abundance increased with increasing CBA, more rapidly in the lower range of CBA

Negative effects of the basal area of conifers within stands on carabid beetles
Although both the total and species-level abundances of forest-dependent species decreased with increasing conifer BA (i.e. decreasing broadleaved tree BA), this response differed among species. The abundance of Pterostichus thunbergi, the dominant species, gradually decreased with conifer BA. This pattern was reflected in the linear decrease of total abundance. By contrast, for the species-level mean response, abundance decreased with increasing conifer BA, and this trend was more rapid in the lower range of conifer BA. This finding suggests that conifers, even in small amounts, have negative impacts on many species, and that natural broadleaved forests are superior habitats for such species. Carabus albrechti albrechti showed such a response to both tree species. In our plantation stands, the lowest proportion of conifers out of total trees was 30% based on BA. An increase in evergreen conifers greatly alters environmental conditions on the ground (e.g. litter, soil, light, and vegetation) in stands where deciduous broadleaved trees dominate, likely leading to reduced habitat suitability for many species that prefer natural broadleaved forests. These negative effects of conifer BA were particularly evident on Sakhalin spruce plantations, although the difference in survey year may have affected this pattern. For Carabus granulatus yezoensis and Synuchus melantho, clear negative responses to conifer BA were found only in the spruce plantations. The crown of Sakhalin spruce easily becomes closed because the whorled branches rarely die and fall (Fujimori 1984(Fujimori , 1985, likely affecting the environment near the ground. Soil was drier and understory vegetation was scarcer in stands with higher Sakhalin spruce BA at our sites (Fig. S6b, d). A previous study on birds in this area also demonstrated a negative response to conifer BA within stands that was clearer in Sakhalin spruce than Todo fir plantations (Yoshii et al. 2015). In the spruce plantations, increasing the amount of conifers within stands may strongly impact environmental factors such as soil dryness, leading to drastic declines of some species.
In Britain, Fuller et al. (2008) compared carabid beetle communities among conifer plantations, mixed plantations of conifers and broadleaved trees, and natural broadleaved forests, and showed that pure broadleaved stands are important habitats for carabid beetles. Similarly, Oxbrough et al. (2012) reported that mixing 15-40% oak trees into Norway spruce plantations was ineffective for carabid beetle conservation in Ireland (but see Skłodowski et al. 2018). In Hokkaido, beetle abundance and species richness in conifer plantations are lower than or comparable to those in natural broadleaved forests (Kaizuka et al. 2020;Yamanaka et al. 2021). Although the habitat function of conifer plantations should not be overlooked, maintenance of broadleaved forests is a reliable conservation method for many carabid beetle species under various conditions.

Implications for forest management and conservation
As strategies to improve plantation multifunctionality in Japan, promoting the recruitment and growth of broadleaved trees through thinning or harvesting and retaining mixed broadleaved trees during thinning and harvesting operations have been proposed and addressed (Yamagawa et al. 2010;Yamaura et al. 2019;Negishi et al. 2020;Forestry Agency 2021). Our results indicate that transforming conifer plantations into mixed forests is insufficient in some cases. Species with lower mobilities or higher dependencies on the environments provided by broadleaved trees would be efficiently conserved by maintaining stands where broadleaved trees dominate. Protecting remnant natural broadleaved forests is likely the best method for conserving species that are not readily conserved in plantations. Alternatively, when conifer plantations with abundant broadleaved trees are regarded as unsuccessful forestry sites, reverting these plantations into broadleaved forests would be effective. Forest management strategies that maintain a certain extent of broadleaved forests are essential to balance timber production and biodiversity conservation.