Co-occurrence variation explains the low host dependence of ambrosia beetles along altitude gradients in SW China


 Background

Separation of biotic and abiotic impacts on species diversity distribution patterns across a significant climatic gradient is a challenge in the study of diversity maintenance mechanisms. The basic task is to reconcile scale-dependent effects of abiotic and biotic processes on species distribution models. However, Eltonian noise hypothesis predicted that the effects of biotic interactions will be averaged out at macroscales, and there are many empirical observations that biotic interactions would constrain species distributions at micro-ecological scales. Here, we used a hierarchical modeling method to detect the host specificities of ambrosia beetles (Scolytinae and Platypodinae) with their dependent tree communities across a steep climatic gradient, which was embedded within a relatively homogenous spatial niche.
Results

Species turnover of both trees and ambrosia beetles have a relatively similar pattern, characterized by the climatic proxy at a regional scale, but not at local scales. This pattern confirmed the Eltonian noise hypothesis wherein emphasis was on influences of macro-climate on local biotic interactions between trees and hosted ambrosia beetle communities, whereas local biotic relations, represented by host specificity dependence, were regionally conserved.
Conclusions

At a confined spatial scale, cross-taxa comparisons of co-occurrence highlighted the importance of the organism’s dispersal. The effects of tree abundance and phylogeny diversity on ambrosia beetle diversity were, to a large extent, indirect, operating via changes in ambrosia beetle abundance through spatial and temporal dynamics of resources distribution. Tree host dependence plays a minor role on the hosted ambrosia beetle community in this concealed wood decomposing interacting system.

experiments are needed to overcome the interactions of different relative importance of biotic and abiotic factors among various spatial scales.
Spatial community structuring in host-speci c niches also implies the presence of demarcated transition zones where host replacement is most likely to occur (Fig. 1, left panel). These 'host turnover zones' are derived from an integral but unspeci ed part of the mutualistic niche concept [17,18]; it can be calculated using beta diversity (hereafter 'βdiversity') weighting [19], as species co-occurrence probably along a speci c environmental gradient. Nevertheless, the niche overlap degree and turnover rate between neighboring spatial communities among different biological assemblages might vary (Fig. 1, right panel). The patterns of β-diversity can be complex and differ among organisms [20], because of simultaneous impacts from extrinsic and intrinsic factors, related to geography and environment and to the evolutionary characteristics of organisms, respectively. Differences in β-diversity occurring along gradients are often used to infer variation in the processes of the structuring spatial communities. At a global scale, the β-diversity of different communities generally decreases along the increasing latitude gradient (from tropical to temperate areas) [21][22][23]. Cross-taxon comparisons through meta-analyses showed a negative, albeit relatively weak, relationship between β-diversity and latitude [20,24]. What's more interesting is that two β-diversity components, including nestedness and turnover which were already proposed by [25], had generally opposing patterns with regard to latitude at a global spatial scale [26]. Recent evidence suggests that gradients in β-diversity occur as a result of latitudinal or elevation gradients and may be caused by a series of combined mechanisms. For example, deterministic processes of environmental or habitat ltering at a regional scale [27] and stochastic processes generating ecological drift at a local scale [28] impact spatial community assembly processes differently. Given the complexity of potential interactions among processes that are not mutually exclusive and are mediated by host speci city, it is not surprising that the relative importance of contributing factors and interactive mechanisms in driving patterns of dissimilarity in hosted biological communities remains largely unresolved.
Separation of the impacts of macro climatic/environmental variables on community dissimilarities remains a challenge that has been rarely explored in the context of large-scale patterns of species richness. In the context of insect communities, we expect substantial compositional changes between regions (if historical bio-geographical effects are important) and along abiotic gradients within regions (if insects generally have speci c abiotic requirements, such as temperature) [29,30]. If insect speciation has been largely induced by allopatric host shifts of specialized insect species, we can expect a complete change in insect assemblage associated with shifts in plant assemblage [31,32]. Comparisons of turnover between plants and the insect community could reveal speciation through host shifts or allopatric speciation with subsequent host specialization. Moreover, these comparisons could also indicate impacts of parallel allopatric speciation or parallel environmental ltering. Thus, it is often di cult to separate patterns induced by parallel responses to macro climatic gradients from those induced by the evolutionary process [12]. Recent studies have attempted to overcome the spatial-scale dependence of diversity through a hierarchically nested method [15,33,34]. A hierarchical modeling framework [35] of β-diversity combined with phylogenetic and functional β-diversity was later proposed and may help to investigate mechanisms of local hosted community assembly. It suggests that if both host plant (henceforth referred as 'host') and insect turnover, which re ect bio-geographical impacts at all scales (from local to regional), occur simultaneously in response to geographic or environmental gradients, we would not expect to nd any positive association between insect community composition and plant species/phylogenetic turnover after accounting for the in uence of geography or the regional species pool. Therefore, if the observed insect turnover patterns were due to host speci city, a close relationship between insect and host species/phylogenetic diversity at both the ne and coarse spatial scales should exist.
Comparatively, if parallel responses of both interacting assemblages to macro abiotic gradients were detected, we could conclude that the turnover patterns would be only correlated at large spatial scales. In this context, recent research has mostly focused on plant-herbivore [12,13,30,36] and -pollinator interactions [37], relationships that are all specialized to some extent. Exceptions are Hulcr et al. [38], Hulcr et al. [1], Wu et al. [39], Wende et al. [40], and Vogel et al. [2], who studied ambrosia and saproxylic beetles, which are loosely associated with their hosts. Their studies investigated the impact of host speci city on beetle community composition within the same climatic region, disregarding environment gradients. Furthermore, the studies aiming to explore the relationships between hosts and insect community along altitudinal or latitudinal gradients at different spatial scales have shown different, and sometimes contradictory, patterns. The interactions of detritivorous insect assemblages with their hosts along steep environmental gradients have rarely been studied, and whether detritivorous insects exhibit similar trends to herbivores and pollinators is still unknown.
Ambrosia beetles (Coleoptera: Curculionidae, Scolytinae, and Platypodinae) normally colonize dying trees; the larvae feed on the symbiotic xylosaprophagous ambrosia fungi that the beetles introduce into the trees [41,42]. Exploitation of the fungi as food source has allowed ambrosia beetles to use a wide variety of hosts [38,43]. The effects of tree diversity, host speci city, and structural heterogeneity of habitat on ambrosia beetles at a local scale and along a at environmental gradient have been well studied [1,2,38,[43][44][45]. However, a regional cross-taxon comparison of how β-diversity of ambrosia beetle communities is related to that of their host species along a steep environmental gradient has not, to our knowledge, been reported. A potential complicating factor in any analysis of diversity gradients (as well as in meta-population studies, species-area analyses, and most other geographically based studies) is that the data are often spatially auto-correlated and scale dependent. It is valuable to use appropriate spatial analysis for testing the degree to which vegetation and insect richness are associated with each other after accounting for the in uence of geography/environment or the regional species pool.
The general goal of this paper was to derive environmental and spatial models to account for patterns of variation in the co-occurrence of ambrosia beetles across a steep elevation gradient in the Yunnan Province, Southwest China. We aimed to compare the relationship between tree's host speci city and ambrosia beetle turnover among different climatic sub-regions embedded within a relatively homogeneous abiotic gradient. We hypothesized that (1) there is a parallel spatial structuring of ambrosia beetles with host trees along the steep elevation gradient, but with a much wider niche overlap compared with those host tree communities, (2) trees' host speci city plays a minor role in structuring ambrosia beetle diversity distribution across both local and regional elevation gradient compared with abiotic environmental factors; and (3) contrasted co-occurrence distribution patterns between host trees and ambrosia beetle assemblages across huge environmental differences explain a weak relationship among these two groups at regional scales, but not for local.

Species turnover comparison between beetles and trees
As Table 1 showed, β-diversity of tree communities measured through Horn similarity method was signi cantly decreased with the elevation increased (GLMs, Z = -5.905, R 2 = 0.0376, P < 0.001) at the regional scale (Table 1) but with a very low R 2 value. Contrary to this, beetles signi cantly increased with elevation gradient (GLMs, Z = 9.295, R 2 = 0.2868, P < 0.001) ( Table 1). Horn similarity β-diversity values of beetle assemblage were signi cantly higher than those of trees in three sub-regions, along the gradient from tropical to temperate (P < 0.0001; Fig. 3a-I). At the local scale, Horn β-diversity of trees and beetles showed a signi cantly different pattern with elevation gradients in tropic, subtropics and temperate respectively, and these opposite pattern was similar with that showed at regional scale (Table 1). Furthermore, most of the Horn similarity β-diversity values of beetle assemblages were statistically signi cant higher than those of trees communities ( Fig. 3a-II, III & IV, P < 0.001), except for values at 2,600 m in the subtropical areas (ns, Fig. 3a-III). When we compared the Horn similarity values of beetles and trees separately, they did not show a uniform tendency neither at regional scale, from tropical to temperate areas nor at local scale, from low to high elevation ( Fig. 3a & Table 1). The observed β-diversity values of tree assemblages measured through Bray dissimilarity method were signi cantly higher than those of beetles (P < 0.01; Fig. 3b). Almost both trees and beetles observed β-diversity values showed a decreased tendency with elevation increased (Z < 0; Table 1) but except Z value of 0.477 for trees community at regional scale. The R 2 values of coe cients between both trees and beetles with elevation gradients also showed a statistically signi cant different pattern at regional and local spatial scales ( Table 1). The abundance β-null deviation measures showed the same pattern with the dissimilarity method through Bray dissimilarity index calculation of tree and beetle assemblages, and most of negative Z values mean a decreased relationship with elevation gradient but except Z value of 0.94 for trees community at regional scale (Table 1; Fig. 3b & c). However, at regional scale, the relationship between β-null deviation of trees and beetles with elevation were signi cantly different with each other (Table 1). If we concentrated this point at local scales, β-null deviation of tree communities were signi cantly decreased with elevation gradient (P < 0.001; Table 1) but beetle assemblages increased. The most of beetles β-null deviation values were statistically signi cant higher than those trees communities at both regional and local scales (P < 0.01; Fig. 3c-I, II&III) but except for those at 3,200 and 3,600 m plots in temperate areas (ns, Fig. 3c-IV).
Correlations of beetle turnover with tree community After removing the in uence of spatial distance, RDA revealed that abundance of the 21 most important tree species of the 211 recorded explained 10% of the variation in wood-boring beetle community composition separately, spatial distance explained 1%, and 12% remained unexplained. Moreover, 14 phylogenetic PC axes (accounting for 90% of variation in tree phylogenetic turnover between sampling units) explained only 7% of the variation in ambrosia beetle community composition separately. At the local scale, the selected tree species abundance after forward selection procedure in RDA analysis separately explained 34%, 3%, and 15% of beetles composition variation in the tropics, subtropics, and temperate area, respectively ( Table 2).

Species distribution ordinations between beetles and trees
As shown in Fig. 4, PCA analyzes showed that all seven selected environmental variables have statistically signi cant cor-relationships with species distribution patterns for both trees and beetles communities (P < 0.05). Among them, the variables of elevation gradients (ELE) and relative annual humidity range (AHR) played a statistically signi cant negative role on species distribution. Considering even all environmental variables were spatial auto correlated with elevation gradient, it's not surprisingly that the minimum temperature of the coldest temperature (MTCM) have a positive impact on species distribution from temperate to tropical areas for both trees and beetles communities (P < 0.05, Fig. 4). More importantly, at the regional scales, the recorded environmental variables had a signi cantly higher explanation power for beetles than trees communities in the present study. It showed that two axis, represented by PCA1 and PCA1 reached to 47% and 24.77% separately for beetles group (Fig. 4-I) but only 17.58% and 16% separately for trees ( Fig. 4-II).

Co-occurrence comparison between beetles and trees
Co-occurrence patterns between beetle and tree community were explored based on a simple comparison of density frequency distribution. As shown in Fig. 5, it showed clear divergent tendency of the curves between beetle and tree communities (P < 0.001, Kruskal-Wallis test). The results also suggested that most ambrosia beetles were collected from the understory, and the highest numbers of Scolytinae and Platypodinae beetles were from the understories at the tropical forest at Bubeng and subtropical forest at Ailaoshan, respectively.

Discussion
Co-occurrences of ambrosia beetles weighted as β-diversity did not all show a statistically signi cant correlation with both trees species and phylogenetic turnover at all three sub-regions from tropical to temperate areas. This result suggests a minor role of host speci city on ambrosia beetle diversity distribution at the community level. Although several studies on comparison of insect host dependence among different climatic regions have presented divergent results [4,5,38,43,46], it is still reasonable to infer a relatively weak biotic interaction between trees and their hosted ambrosia beetles when comparing the impacts of huge climatic differences from tropical to temperate climatic gradients. That is, elevation gradients, the most important environmental variable, are major drivers which structuralized the species distribution patterns at regional scale for both beetles and trees.
As the ENH explains, Eltonian factors correlate closely with scenopoetic variables at the regional scale, which thus capture an important part of the biotic signature for ambrosia beetles [47,48]; however, factors such as host speci city dependence may not affect distributions at the regional scale of large extents [35]. Eltonian's ecological niche model has been observed through a series of parasitically, competitively, and antagonistically interacted organisms across huge abiotic environment difference [34,47,49]. Nevertheless, this model has been questioned and a different hypothesis was also proposed recently by several other studies [50,51] and debates arose as to whether biotic interactions exert a dominant role in governing species distributions at macro-ecological scales. It emphasized that some explicitly incorporated biotic variables could generally improve species distribution models and have independent contributions [51]. As far as we know, this inconsistency relating to impacts of biotic interaction on species distribution models at macro spatial scales are not prevalent, although a most recent systematical analysis found that the spatial scaling patterns of β-diversity are consistent across metrics and taxa [52]. Moreover, it was probably mediated by an interaction that was context dependent and considering spatially biased environmental variables. Therefore, our study supports the effectiveness of the ENH. This hypothesis explained an approximately parallel diversity distribution pattern between trees and their hosted ambrosia beetle community across macro climatic gradients, but not for a local scale.
Admission of the minor role that insect-plant interactions play on large-scale hosted species turnover does not imply denial of the associations between different nested trophic assemblages. Tree host speci city was an important topic in biodiversity estimation and maintenance at local scale, although it is di cult to measure host speci cities directly.
First, comparing the strength of tree host speci cities among different climatic areas generally ignore the impacts of the species pool and has led to diverse interpretations [5,43,46]. Second, simply using tree diversity as a surrogate to estimate their hosted herbivorous insects diversity [3,4,13,38,39,47] is too coarse to get convincing unbiased results when considering the mixed effects of spatial environmental variables. Most of the different insect assemblages have shown a loose relationship with their hosts, and host shifting is a common phenomenon in nature [53]. If we further explore mutualistic interactions in detail, most of these interactions present a broad range of partners, even if it allows for ineffective partners to persist [18]. For ambrosia beetles, as a major group of wood decomposers, a large part of this taxon has a strong dependence on their mutualistic fungus. It means external biotic selection pressure caused by the tree host speci city might be minor compared with other factors. However, researches emphasizing the role of host speci city in insect diversity had shown different results. For example, Whitfeld et al. [13] concluded that plant phylogeny could not explain much of the variation in herbivore abundance, but exudates and leaf nitrogen content had a strong impact on caterpillar abundance. This result illustrated the process of conserved traits at one trophic level to in uence community-wide patterns at another. Patterns that contradict each other even within the same region have also been observed in the Cape Floristic Region of South Africa [10,11,54] as aforementioned. Current studies suggest that extremely specialized or generalized insect species communities, whether detritivores or herbivores, only account for a minor proportion of the total assemblages, and most species are concentrated in the center group of the host dependence sequence [4,5]. In this context, ambrosia beetle diversity distribution was also rarely impacted by tree host speci cities at the local spatial scale, especially when considering that limited host dependence will decrease quickly during the wood decomposition process [39].
Additionally, biotic and abiotic factors and interactions of ambrosia beetle diversity distribution across a large climatic gradient impacts species dispersal ability. This dispersal ability, represented as access to dispersal or colonization over a relevant time interval, was another important component to mediate the actual spatial area of distribution of the species [48]. It is possibly that co-occurrences of interacted assemblages vary widely among organisms with different dispersal abilities [20,24,26], when we did cross-taxon's β-diversity comparison across a same environmental gradient. The present study showed a clear pattern that tree dispersal via seeds had a much higher turnover than the actively mobile ambrosia beetles. It con rmed that species characteristics, re ecting the autecology of individual organisms, have a statistically signi cant effect on β-diversity distribution. Because the high dispersal ability signi cantly reduced the plot sampling difference, it should have a negative impact on β-diversity.
Nonetheless, this is not what we found. Harrison et al. [55] compared cross-taxon' β-diversity with different dispersal ability for British biota ranging from ying birds to insects and plants. They showed that β-diversity is more constrained by species' ecological niche requirements than by the species dispersal ability; and it could be negligible for statistical model comparison. The organisms in this study have huge differences in dispersal ability, but the βdiversity of a speci c biological assemblage might be controlled through a series of combined and interacting factors at different spatial and temporal scales. Species dispersal ability can also re ect species environment adaption exibility to some extent. When species both adapt and disperse at the same time interval, dispersal and adaptation cannot combine positively to affect biodiversity maintenance [56]. Such con icts can be observed and it gives us an idea that species diversity distribution is a paradox of mixed factors that cannot be easily and clearly separated. Although it is valuable to get a deep insight for comparisons of dispersal and adaptation among different biotic assemblage, our study presented a statistically signi cant pattern of β-diversity between trees and their hosted ambrosia beetles attributed most probably to those dispersal ability and type variation.
As we try to explore the major underlying mechanism that mediated the relationships between trees and the hosted ambrosia beetles, many other technical factors, including effectiveness of sample size, spatial grain of sampling, bias of statistical method used, and data weight of each biological individual, can mix patterns that are not interpreted easily. Bennett & Gilbert [57] compared the differences between multi-site dissimilarity and traditional null model methods, and Engel et al. [58] used the coverage-based rarefaction method to overcome the species pool dependence of β-diversity bias toward to null hypothesis. No matter what kind of speci c method is used nor how many effective samples were adopted, it is de nitely a challenge to arrive at a convincing explanation based on statistical inference directly. Considering that although from a human perspective of the "local" scale, is standardized to be de ned by 20×25 m plots in the present study, from the "point of view" of beetles and trees, these plots are not "equally local", and the way these taxa perceive the environmental heterogeneity at this spatial scale could differ greatly just because of differences in their body size, which might impact on the human perception of their species turnover patterns. Furthermore, when we look back to compare the basic physiological differences between trees and beetles, tree distribution depends more heavily on soil water and nutrient content than do beetles. Both assemblages faced the same temperature gradient at a macro scale; ambrosia beetles clearly have a broader low-temperature tolerance ability compared with that of tree communities because the majority of the beetle's life cycle is con ned to within the decaying tree stem. That is, if we did comparisons with some other insect assemblages which are not con ned to wood boring ecological system, for examples, ants community [59](Castro et al., 2020) and herbivorous insects [60] in mountainous areas of Neotropical area, β-diversity of latter two groups, and which were mainly generated through turnover rate, were strongly in uenced by variables correlated with elevation gradient, habitat structure and local resource distribution. From the data of species co-occurrence distribution in the present study, some of the beetle species collected in tropical Bubeng were also found in subtropical Ailaoshan. It suggests that both host dependence and temperature gradient from tropical to subtropical areas do not have a statistically signi cant impact on ambrosia beetle distribution. Thus, the parallel co-occurrence distribution pattern between ambrosia beetle and tree assemblages at regional scale are most probably attributed to macro climatic gradient, and without relationships to trees host speci cities clearly.
Finally, as previous study demonstrated that ambrosia beetles rely on ethanol for host tree colonization because it promotes the growth of their fungal gardens while inhibiting the growth of "weedy" fungal competitors [61], it seems explained the most of ambrosia beetle specimens were collected at the understory FITs in tropical and subtropical sampling areas. Because the tree stems at understory are generally relative bigger and older than canopies' woods and can emit micro dose ethanol out from bark within dim light environment. The 75% alcohol used here might has in uenced the community composition dramatically, making the ethanol-attracted component such as the genera of Scolytoplatypus and Xylosandrus are much more prevalent in our collections. As regarded to statistically nonsigni cant differences of ambrosia beetles collected between canopy and understory FITs in cold temperate, it was most probably attributed to the relative lower density of wood stems over there and similar light conditions among them.

Conclusions
In this study, we compared co-variations in β-diversity of wood-boring ambrosia beetles and their host tree communities from the tropics to cold subalpine habitats in Yunnan, SW China. The results showed that species turnover of both trees and ambrosia beetles have a relatively similar decreased tendency along the climatic proxy (represented by the elevation gradients) at the regional scale, but not at local spatial scales. This pattern further supported the ENH according to which emphasis is on macro-climate in uences on local biotic interactions between trees and hosted beetle communities, whereas local biotic relations represented by host speci cities dependence are regionally conserved. At a con ned spatial scale, cross-taxa comparisons of β-diversity highlighted the importance of the organism's dispersal ability and type which was most probably inherited by characteristics of speci c biological groups. The effects of diversity of tree abundance and phylogeny on ambrosia beetle diversity were to a large extent indirect, operating via changes in beetle abundance through spatial-temporal dynamics of resource distribution. Tree host dependence, which was considered and represented by host speci cities, plays a minor role on the hosted beetle community in this concealed wood decomposing interacting system.

Forest plots from tropics to temperate
The study was performed in the Yunnan Province, Southwest China, located centrally to the north of the Indo-China Peninsular (Fig. 2, left panel). The area is extremely diversi ed in habitats and biodiversity, and it is one of the 25 biodiversity hotspots in the world [62]. The region sampled in this study is also part of the 'Eastern Himalaya-SE Tibet hotspots', and area with up to 3,000-5,000 or more vascular plants per 10,000 km 2 [63,64]. The impacts of the tropical monsoonal climate and complex mountainous topography cause this area to be covered with diversi ed vegetation, from tropical monsoonal rainforest to temperate coniferous forest, along elevation and altitudinal gradients.
We sampled ambrosia beetles from three typical biomes: tropical monsoonal rainforest in Bubeng (Xishuangbanna); subtropical mid-mountain moist evergreen broad-leaved forest at Ailaoshan (Puer); and cold temperate spruce-r forest at Yulongxueshan (Lijiang) from low to high elevation (Fig. 2, right panel). Three big permanent plots for longterm ecological research with areas of 20 or 25 ha have been established in recent years, and a standardized vegetation inventory has shown that the area at Bubeng is hyper-diversi ed [468 woody species, [65]], the Ailaoshan plot is median-diversi ed [103 woody species, [66]], and the coniferous forest at Yulongxueshan is low-diversi ed [62 woody species, [67]]. Basic environmental information of each sampling plot are listed in Table S1.

Sampling design
A hierarchical and spatially nested sampling approach was used. To compare the impact of elevation on the βdiversity of ambrosia beetle species across the three different sub-regions, we established three elevational transects in each sub-region, close to the permanent plots aforementioned. The experiment was performed from March 2018 to May 2019. In each area, we studied the same ecological amplitude of elevation gradient and covered a range of almost 400 elevation meters. Each transect included ve forest plots with an intermediate cell size of 25 × 20 m, all of which were equipped with devices for collecting beetles. All ve forest plots at each transect were oriented parallel to the respective contour lines and were at least 40 m apart. That is, each sub-region had fteen cell sizes of 25 × 20 m plots and three sub-regions from tropical to temperate climate totally included 45 sampling plots of vegetation.
The tropical transects were located in Bubeng, Xishuangbanna (21°61 N, 101°58 S). This area borders Myanmar in the southwest and Laos in the southeast. Average annual temperature and precipitation are 22°C and 1,500 mm, respectively. The rainy season occurs from May to October and the dry season from November to April, with approximately 80% of the annual precipitation occurring in the rainy season. Heavy fog frequently occurs in the lowlands and valleys in the mornings during the dry season, and this fog expands the northern limit of the tropical rain forest from Southeast Asia. Three separated elevation transects (at 600, 800, and 1,000 m) were established, based on vegetation and topography.
The subtropical transects were established in Ailaoshan, Puer (24°53 N, 101°03 S). Average annual temperature and precipitation are 11°C and 1,900 mm, respectively. The dry season occurs from December to April and the rainy season from June to October. The area encompasses a large tract of evergreen broad-leaved forests primarily dominated by Lithocarpus and Castanopsis species at mid-elevation (ca. 2,200-2,600 m a.s.l.) with a dense or sparse understory of bamboo and with Rhododendron dwarf forest toward higher elevations. Three separated elevation transects (at 2,200, 2,400, and 2,600 m) were established based on their vegetation and topography.
The cold temperate transects were established at Yulongxueshan, Lijiang (27°14 N, 100°23 S). Average annual temperature and precipitation are 5.5°C and 1,600 mm, respectively. The rainy season occurs from July to September and the dry season from December to March of coming next year. The area encompasses a large tract of temperate coniferous forest with an understory formed primarily with Berberidaceae, Caprifoliaceae, and Rosaceae species. Trees of the families Pinaceae and Fagaceae dominate the canopy equally. Three separated elevation transects were established (at 3,200, 3,400, and 3,600 m).

Insect sampling and diversity estimation
Beetle sampling was performed using modi ed aerial collectors both in the canopy and understory of each forest plot at all sites. Aerial collectors, also called ight interception traps (FITs), were constructed with two hard transparent plastic plates (50 × 35 cm; H × W), which were arranged crosswise and xed upon a plastic bowl (35 × 30 cm; D × H). A round plate of soft transparent plastic (45-cm diameter) was xed over each FIT as a roof to prevent too much precipitation entering the trap during the rainy season. Within each plot, one trap was installed on tree branches in the canopy, at a height of 10-30 m above the ground, and a second was placed in the understory area at a height of 2 m. The FITs were xed with nylon ropes to prevent the spillover of anti-rotting liquids by the wind. The collection bowls of the FITs were lled with a mixture of 75% alcohol and blue colored anti-freeze (ethylene-glycol) at a proportion of 1:2 v:v. Ten FITs were used in each transect, and a total of 90 FITs were installed across all three investigated regions.
Because of the large differences in climate among the three regions, the trap collection activity started at different times in each region; however, the sampling always covered the period of peak beetle activity and lasted at least one whole year. We conducted eld work at Bubeng, from the beginning of April 2018 to the end of March 2019; at Ailaoshan, from the beginning of May 2018 to the end of April 2019; and at Yulongxueshan, from the beginning of June 2018 to the end of May 2019. At all plots, traps were emptied every 10 days during the collection period (with a few exceptions, where the traps were destroyed by strong winds or collection was impossible because of heavy rains). The beetle specimens were preserved in 70% ethanol and were preliminary sorted to morph-species level in the laboratory. Later all morph-species were further determined by Prof. Roger Beaver in Thailand and Dr. Heiko Gebhardt in Germany to the species level. Data analyses were based on numbers of species and individuals combined from all trap times per plot and the total counts from all collection periods. Voucher specimens of the collected beetles were temporarily stored in the laboratory at the Honghe University.
Both multi-sites dissimilarity and null model methods were used to calculate the β-diversity of beetle's community.
The rst method used the package 'vegetarian' [68] in R 3.6.3 (R Core Team, 2013). The Horn similarity index was used as recommended by Jost [69], because it was the only overlap measure that was not disproportionately biased toward rare or common species. This index is considered to be a "true" measure that quanti es effective species overlap between sampling units [70]. The Horn similarity index was de ned as: 1 D β = (ln2 -H βShan ) / ln2, in which H βShan is the Shannon entropy based on Hill numbers and β-diversity is therefore independent of α-diversity [69]. To overcome the demographic stochasticity of the sampling plots, a null model method was used to simulate species assemblages for each FIT by randomly sampling individuals from the regional species pool, according to the relative species abundance in the regional pool and the total number of individuals [71][72][73]. The dissimilarity matrix was calculated using the Bray-Curtis method, which takes account of species relative abundances [70]. From 1,000 iterations of the null model, we calculated a standardized effect size (β-deviation) as the difference between the observed and mean expected dissimilarity, divided by the standard deviation of expected values. This formula retains the observed abundance distribution but randomizes the location of sampled individuals. The R code for β-null deviation calculations referenced the methods provided by Stegen et al. [74]and Tucker et al. [75].
The canopy and understory FITs were combined to give the smallest sampling unit for the estimation of β-diversity among three regions. Pairwise Horn similarities were calculated between all plots from beetle species abundance lists compiled for each sampling plot. The function 'sim.table' in 'vegetarian' was used for this purpose. Given that we were computing and comparing turnover between identical sampling units in all cases, it was not necessary to consider the species accumulation curve to assess whether sampling was adequate.
Tree sampling and diversity estimation The vegetation data collection was performed in April and May of 2019. The ve 25 × 20 m forest plots established for the insect sampling at the three study regions were also used for the vegetation survey. Identical eld methods were used to survey each plot. All plots were established as far away as possible from the large canopy gaps created by recent anthropogenic and natural disturbances. In each plot, we measured the abundance of each tree species (or morphospecies) ≥ 5.0 cm diameter at breast height (1.3 m). Seedlings were excluded. All sampling methods used in present study comply with the instructions of the Center for Tropical Forest Science (CTFS; http://www.ctfs.si.edu/) for collection of long-term, large-scale forest data from the tropics [76] and with those from the Chinese Forest Biodiversity Monitoring Network (http://www.cfbiodiv.org/). When establishing plots on slopes, we positioned the plot center line perpendicular to the slopes to minimize elevation gradients within the plots. As sampling included similar numbers of plots spanning small and large spatial distances, we were able to compare the potential in uence of spatial limitation between regions at similar scales, including extents that encompass typical dispersal distances (seed shadows) from tropical to temperate vegetation.
Tree abundances were recorded for each plot. Horn similarity matrices were constructed using plots and transects as sampling units in the same manner as described for the beetles. The family and genus names of all the studied species (216 species in total) in the APG III system were obtained using the R package 'plantlist' [77], and their phylogenetic relationships were examined using the online Phylomatic tool ( [78]; www.phylodiversity.net/phylomatic/) based on the angiosperm consensus tree from Davies et al. [79]. Similarity matrices were then constructed for plant phylogenetic β-diversity [PhyloSor Index [80]] with the function 'phylosor' in 'picante' package in R [81]. PhyloSor is a modi cation of the Sørensen similarity index that quanti es phylogenetic similarity of communities as the proportion of shared phylogenetic branch-lengths between two samples. If the length of shared branches is high, communities comprise phylogenetically closely related taxa.

Environmental factors
We recorded air temperature and humidity data at each of the transects within the three study regions every 30 min using a thermo-logger (DS1923Hygrochron iButton®, Maxim, CA, USA) from April 2018 to May 2019, during the period of insect collection. A total of nine environmental data loggers were used; each device was xed at one of the ve canopy FITs in each transect. A total of seven variables were measured, including annual mean temperature (AMT) and humidity (AMH), annual temperature (ATR) and humidity ranges (AHR), maximum temperature of the warmest month (MTWM), minimum temperature of the coldest month (MTCM), and average elevation (ELE) of each transect. These data were assembled into a secondary environmental matrix and were prepared for canonical redundancy analysis (RDA). Detailed data are given in Table S1.

Data analysis
Spatial scale of species turnover We rst compared both beetle and tree composition turnover at local and regional scales using the multi-site dissimilarity method. A grouped plot-level similarity matrix was calculated and then partitioned into two different independent spatial components that re ected various β-diversity levels. It included turnover between sampling plots (total 15 plots) within three sub-regions and turnover between sampling plots among three sub-regions (total 45 plots). Considering all these turnover values produced by the multi-site dissimilarity method fall in the range from minimum zero to maximum one, beta regression method in Generalized linear models (GLMs) was performed to detect the relationships between both beetles and trees composition with variable of elevation gradient respectively. Then a Nonparametric Kruskal-Wallis ANOVA (analysis of variance) was further conducted to test for differences in Horn similarity values at various group and two spatial scales, followed by the appropriate post hoc tests. Data from the whole year of collections were combined for these analyses. The spatial component of turnover in tree species composition was investigated in an identical manner.
We also examined how β-null deviation values changed for meta-communities along the gradient from tropical to temperate for both tree and beetle community through beta regression method in Generalized linear models (GLMs) as mentioned above. The difference, in units of standard deviations, between the observed and mean expected raw turnover provided a measure of value that had sampling effects removed. The β-null deviation values based turnover estimates between tree and beetle assemblages are directly comparable to each other using a nonparametric Wilcoxon paired test, and any remaining correlation they had with gamma diversity (or other explanatory variables) could be interpreted as evidence for non-random ecological processes leading to intra-speci c aggregation [82].

Correlations of beetle turnover patterns
A correlation (RDA) approach was used to test for association of tree and beetle species composition while controlling for spatial distance between sampling plots. First, we used forward and backward selection in an RDA assessing the in uence of the abundance of individual tree species (Hellinger transformed) on beetle species composition to identify the most important tree species for inclusion in the ordination. This was necessary because there were more tree species than sampling units in our dataset. The selection procedure was conducted using the 'ordistep' function in 'vegan' package. RDA ('rda' in 'vegan') was then performed on beetle species composition (Hellinger transformed) with the selected tree species as constraining variables and spatial distance [converted to a rectangular matrix using PCNM [83], the 'pcnm' function in 'vegan'] as the conditioning variable (i.e., the effect which is removed). After that, variation partition was used to quantify the relative importance of spatial distance and trees abundance in determining the accompanying beetle species composition with an 'anova.cca' test in vegan. Similarly, to assess the impact of plant phylogenetic composition on beetle species composition, we rst used phylogenetic principal components analysis ('phyl.pca' in 'phytools') to select the set of PC axes that explained 90% of variance in plant phylogenetic community composition. RDA was then performed on beetle species composition (Hellinger transformed) with selected PCNM converted spatial distance as the conditioning variable. P-values were assessed based on 999 random permutations. Because we wanted to explore the relationship between the dissimilarity of communities with environmental factors and spatial distance, we log-normalized the explanatory variables to make them comparable and then converted them to separated distance matrices. If both tree and beetle turnover occurred in response to climatic gradients or re ected bio-geographical in uences (regional scale), we would not expect to nd any positive association between beetle species composition and plant species/phylogenetic turnover after accounting for the in uence of geography (local scale).

Coordination of beetles and trees
To further compare the differences of relationship between beetles and trees with environmental variables at regional spatial scale, principal component analysis (PCA) was applied to the environmental variables, and the statistically signi cant components were selected by RDA. Beetle and tree species composition data were also Hellinger transformed. To quantify the homogeneity of dissimilarity variances within each transect, we compared the variances in the dissimilarity matrix using the 'betadisper' method [84]. This test is analogous to Levene's test for homogeneity of ANOVA variances.
Comparsions of co-occurrences frequency density between beetles and trees Finally, we used all of similarity values which were produced through Horn similarity methods to compare the community compositional differences between ambrosia beetles and trees along the elevation and spatial scales in the present study. It provided a straightforward observation on species turnover differences between ambrosia beetles with their hosted trees communities.  Figure 1 The space of a host can be subdivided into separate sections, each depending on the presence of particular symbionts (1 and 2) with distinct optimum performance along the respective niche axes (i.e., the elevation niche). Left panel is re-sketched from the gure 1 of Rolshausen et al. (2020). Right panel illustrates the niche overlap difference among different taxon across the same environmental elevation gradients.  Turnover (Horn similarity) of insect species (ambrosia beetles, yellow) and tree species (green) between Flight intercept trap sampling plots (25 x 20 m) at one regional scale (I) and three local scales (II, III & IV) in Yunnan

Figures
Province. Fig.3b shows the observed β-diversity calculated through Bray-Curtis dissimilarity method. Fig. 3c shows the values of β-deviations through null model method which is based on the differences between observed β-diversity and a standardized effect size of β-diversity that controls sampling from the regional species pool. Boxes represent the median and 25th/75th percentile, and whiskers extend to 1.5 times the interquartile range. PCA analysis of the correlation between communities of ambrosia beetles (I) and trees (II) with selected seven environmental variables at regional scale based on the RDA method. All ambrosia beetles and trees abundances data were Hellinger transformed. Figure 5