Assembly Processes Structuring Mammal Metacommunities are not Zero-sum

Assembly Processes Structuring Mammal Metacommunities are not Zero-sum | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL This is a preprint and has not been peer reviewed. Data may be preliminary. 22 September 2025 V1 Latest version Share on Assembly Processes Structuring Mammal Metacommunities are not Zero-sum Authors : Baoxiang Huang 0000-0003-3076-3239 , Marcel Holyoak 0000-0001-9727-3627 , Nathan Roberts 0000-0001-8299-4499 , Dongqi Liu , Wannian Cheng , Dusu Wen , Jingxuan Wang , Yanhui Guan , Heng Bao , Jinzhe Qi , and Guangshun Jiang 0000-0001-6321-9489 [email protected] Authors Info & Affiliations https://doi.org/10.22541/au.175856347.70125992/v1 234 views 147 downloads Contents Abstract Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Metacommunity studies commonly gain insight by partitioning assembly processes into habitat filtering, dispersal and species interactions. These processes are commonly assumed to be zero-sum, such that structuring by one process leads to less structuring by other processes. We predicted that mammal community composition would be strongly affected by species interactions in geographical areas with abundant top predators, while dispersal and habitat selection would be stronger in areas lacking top predators. Unexpectedly, we found that in mammal communities in Northeast China forests, strong environmental/habitat filtering was coincident with strong effects of top-predators (species interactions) on other community members and marked roles of dispersal. In areas with few top predators and less habitat/environmental filtering, species interactions remained important but driven by mesopredators rather than top predators. Multiple types of analysis support our finding that strong habitat structuring of mammal communities can be accompanied by strong effects of species interactions and dispersal. 1 Introduction Studies of metacommunities have progressed from early paradigms or perspectives (Leibold et al. 2004) to embrace a variety of frameworks and methods, such as stochastic vs. deterministic processes, niche vs. neutral and variation partitioning methods (Brown et al. 2017). Despite a plethora and confusion of methods, we have found one framework to be particularly effective for understanding metacommunity processes (Holyoak et al. 2020), partitioning community composition into the processes of habitat filtering, species interactions, and dispersal processes (Biswas & Wagner 2012; Marrec et al. 2018; Holyoak et al. 2020). An important assumption of this method, like variation partitioning analysis (Cottenie 2005; Brown et al. 2017), is that community composition or processes are zero-sum, such that more of one process requires a reduction in another process. Surprisingly, this assumption has not (to our knowledge) been examined, and we will show that doing so provides substantial insights into metacommunity dynamics. In natural communities it can be difficult to disentangle the roles of different assembly processes, especially for large animals that are difficult to manipulate. For instance, most studies of mammal community assembly are correlational, providing weak inference (Brown et al. 2017). A way to move beyond correlational studies is to utilize large geographic areas that are relatively static in the presence or absence (or abundance) of species such as top predators over long periods of time, which creates an opportunity for natural experiments comparing forests with or without of particular predators (Liu et al. 2025). Top predators are known to be important because of their strong interactions with other species and effects on mammal community structure (Soulé et al. 2005; She et al. 2025). A combination of top predators and bottom-up effects set the vertical structure of mammal communities and play substantial roles in maintaining mammal biodiversity (Johnson & Fisher 2007; Letnic et al. 2009; She et al. 2024). Here we use such differences in the presence and abundance of top predators in large areas of forest to explore the balance between habitat filtering, dispersal and species interactions in structuring mammal communities. We also take advantage of a major highway as a potential dispersal barrier to test inferences about the role of dispersal. Forest mammal communities are important indicators of global change (Selva et al. 2024) and are also undergoing changes in habitat distributional patterns (O’Connell & Hallett 2019) and species interactions (Tylianakis et al. 2008). For example, habitat succession has substantially changed small mammal communities in Glacier Bay, Alaska (Sytsma et al. 2023). Distributional changes and other responses to land-use and climate change also potentially alter species interactions, yet they are difficult to study (Tylianakis et al. 2008). Changes in species composition are the most direct reason for changes in species interactions (Spiesman & Inouye 2013). Dispersal is also expected to affect the organization of species in interaction webs (Thompson & Gonzalez 2017). In the current study, we take advantage of large spatial scales at which changes in the distribution of top predators are slow, so that top-predator distributions are relatively static within the 3-year period spanned by our study. Existing studies generally recognize that both deterministic and stochastic processes drive terrestrial community assembly (Jiao et al. 2022; Riddley et al. 2025). It is also recognized that dispersal and niche processes (habitat filtering and species interactions) can have both deterministic and stochastic components (e.g. (Brown et al. 2017). For local communities, deterministic processes mainly include habitat selection/filtering and species interactions (Chen et al. 2024), whereas purely stochastic processes mainly reflect random variation of species abundance and associated variation in dispersal (Chase & Myers 2011). Neutral Community Models (NCM) that consider population dynamics but not species’ evolution are useful tools for investigating assembly processes (Sloan et al. 2006). Predictions of the relationship between relative abundance and occurrence frequency in NCM’s can be used to quantify the processes of deterministic versus stochastic dynamics. Occurrence and abundance are expected to increase with dispersal rate, and such techniques have been widely used to elucidate the ecological process of complex communities (Burns et al. 2016; Cheng et al. 2024a). Fitting NCM’s to community data provides estimates of per capita movement rate, m , and how much variation can be accounted for by purely neutral processes. Prior studies in our Northeast China study area have documented the roles of habitat and environmental factors, and roles of top predators in top-down effects at a functional group level (She et al. 2023). They have not investigated the role of habitat filtering, dispersal and species interactions in community assembly processes. In large forest areas top predators are relatively static in their abundance and presence/absence relative to the timescale of our study (Li et al. 2024). The forest structure, long-term top-predator abundances and trajectories in abundance of herbivore species are largely a result reductions in human activities over a 27-year period (1998 on) as people have moved out of forest areas (Jiang et al. 2017). We addressed two questions in northeast China forest mammal communities: 1) How do habitat (including environmental) factors, species interactions and dispersal influence mammal (meta)community composition across sites that differ in the presence/abundance of top predator species? 2) Are the assembly processes of habitat filtering, dispersal and species interactions zero-sum or not? We utilize natural experiments comparing top-predator-abundant forest (TPAF) and top predator low-abundance forest (TPLF), locations with and without particular top-predator species, and on either side of a potential dispersal barrier. In what follows, we first describe how top-predator abundance alters the background patterns of species richness, beta diversity and composition (using analysis of similarity, ANOSIM, and similarity percentage, SIMPER). We then investigated changes in community similarity over space using distance decay relationships (DDR) to learn more about dispersal in the presence and absence of a highway that will be shown to act as a barrier to dispersal. Next, we used Variance Partitioning Analysis and redundancy analysis (RDA) to investigate habitat associations/filtering in TPAF vs. TPLF. Lastly, we examined the fit of neutral community models (NCMs) to communities with high or low predator abundance and forest areas with and without particular predators to understand the role of species interactions in modifying NCM fit; this allows us to test the role stochastic factors have in determining species relative abundance and frequency of occurrence in local communities and to estimate movement rates. Overall, our analyses provide surprising insights into how top predators affect the structure and assembly of forest mammal metacommunities. 2 Methods 2.1 Study Area We selected two study areas which are both Amur tiger habitats in temperate forests (She et al. 2023). Based on detected occupancy and abundance over 3 years, Hunchun forest has abundant top predators (n=82 camera traps) and for reasons justified in the Results we call it top-predator abundant forest (TPAF), while top predators in Xiaobeihu forest were less frequently detected (top predator less-abundant Forest, TPLF n=78, Table S2). Of these two study areas, the Xiaobeihu forest in the northwest is located in the Xiaobeihu National Nature Reserve, Heilongjiang Province (44°3–44°18’ N, 128°33–128°45’ E), which is part of the UNESCO-approved Jingpohu National Geopark. Vegetation in the reserve is varied, with volcanic crater forest being the dominant vegetation type. Protected rare mammals include Amur tiger (Endangered, EN) and Asiatic black bear (Vulnerable, VU). Hunchun forest is located in the Jilin Hunchun Amur tiger National Nature Reserve next to the China-Russia border (42°24’–43°28’ N, 130°14’–131°14’ E; Figure S1). Many rare mammals occur in the reserve, including Amur leopard (Critically Endangered, CR), Amur tiger (EN), Siberian musk deer (VU), and Asian black bear (VU). This area is the main habitat of Amur tiger and leopard in China (Wang et al. 2023; Jin et al. 2025). Hunchun forest is bisected by the 331 National highway, and we refer to the China side, TPAF-C (top-predator abundant forest China side, n=43) and Russia side, TPAF-R (top-predator abundant-forest Russia side, n=39). While both sides are inside China, the latter is closer to the Russian border. Hunchun also has a high incidence of human-wildlife conflicts due to an elevated number of human disturbances (Yang et al. 2019; Cheng et al. 2024b). 2.2 Camera Trapping We designed a 20×20-km monitoring grid composed of 100 grid cells in each of the two study areas and ran one camera trap at the center of each grid cell from October 2019 to October 2022 (Figure S1; (Dickman et al. 2025). The wide spacing of cameras reduces spatial autocorrelation among camera trap locations (She et al. 2024). This study analyzed only warm-season (leaf phenology from May to October in Hunchun) mammal communities to avoid the added complexity of seasonal changes (Cao et al. 2022). To ensure comparability across camera-trap stations, we retained only camera locations that were functional for ≥90 days (Xiao et al. 2019). We maintained the cameras according to standardized operational requirements (Wen et al. 2024), recording geographical coordinates and independent activity events of mammal species and human disturbance (i.e., people, cattle, and domestic dogs). Detections of the same species within 30 minutes were considered to represent a single event (O’Brien et al. 2003). A total of 21 forest mammal taxa were detected (Table S1). Taxa were generally species; however, because it was difficult to identify species of small nocturnal rodents using camera-trap images, we lumped small rodent species into a single taxon called “mice”. The Relative Abundance Index (RAI) estimates the number of independent individuals per species recorded within 100 days per camera trap. There is good evidence that for tigers camera trapping gives an RAI that is related to tiger density (O’Brien et al. 2003). We used all RAI values to calculate community composition at each camera-trap location. 2.3 Habitat Metrics and Analyses For habitat variables, 1) we recorded the diameter at breast height (DBH), forest type (six types: birch, larch, oak, coniferous, broad-leaved/deciduous, coniferous-and-broad-leaved-mixed), species and density of trees (>2 m DBH) within a 30×30-m sample plot in front of each camera. 2) We obtained Landsat8 data for Hunchun and Xiaobeihu forests to calculate Normalized Difference Vegetation Index (NDVI), and downloaded ASTER GDEM 30-m grid data from Geospatial Data Cloud site to extract elevation, slope and aspect of camera trap locations (GDCs 2025). 3) We obtained vectorized data for villages, roads, rivers and lakes in Hunchun and Xiaobeihu forests from Chinese National Catalogue Service for Geographic Information (CNCSGI 2025). Prior to testing the relationship between environmental factors and differences in mammal community composition, we ensured the independence of habitat variables (Spearman’s ρ <0.5, Figure S2). We then used Variance Partitioning Analysis (VPA) to test which environmental factors were correlated with differences in mammal community composition (Minchin 1987; Borcard et al. 1992; Legendre & Legendre 2012). Secondly, Redundancy Analysis (RDA) was used to analyze the environmental and geographical factors that affect the differences of mammal community composition between TPAF and TPLF (Legendre & Legendre 2012). Finally, Distance Decay Relationships were analyzed using linear regression to evaluate the Bray-Curtis similarity of community composition between camera traps vs geographical distance in different forest areas; because the data are not independent, we resampled the dataset (n=1000), bootstrapping 1000 times to calculate significance and 95% confidence limits (Dvison et al. 1986). 2.4 Diversity Analyses We used observed species richness and Shannon diversity ( H’ ) to calculate α diversity of the mammal communities as , where P i is the proportional abundance of species i . To quantify β diversity, we applied Permutation Multivariate Analysis of Variance (PERMANOVA) to test the significant difference in species community composition between TPAF and TPLF (Warton et al. 2012). We then used Analysis of Similarities (ANOSIM) to test the difference in beta diversity between multiple groups of mammal communities. Lastly, we used Similarity Percentages (SIMPER) to test for between-community differences in RAI for each species, and evaluate the percent contributions of each species to the average between-area Bray-Curtis dissimilarity (Clarke 1993). 2.5 Community Assembly Analyses We used Neutral Community Models (NCM’s) to quantify the role of stochastic processes in community assembly (Sloan et al. 2006). This method predicts the relationship between the occurrence frequency and RAI of each species in a community. The species were divided into three groups based on their occurrence frequency and 95% confidence intervals predicted by the model: above prediction, below prediction, and not different from a neutral process at p <0.05 based on the overlap of 95% confidence interval with the NCM-predicted value (Zhou & Wang 2023). Occurrence frequency of a species below neutral-model-predicted values for a given RAI in mammal communities reflects things like habitat specialization, negative effects of species interactions or dispersal limitation. To further test the role of top predators in the composition processes of mammal communities (community assembly), we removed locations containing recorded occurrence of the four top predators (Amur tiger, Amur leopard, Asian black bear, brown bear) or sable, and compared the NCM changes of each mammal population in the remaining “no-top-predator” or ”no-sable” locations to judge the effect of each predator species on community assembly. We also used the R 2 value to assess fit of the stochastic NCM and calculated the proportion of species that were not neutral (above or below prediction). We resampled the dataset (n=1000), bootstrapping 1000 times to calculate 95% confidence limits for coefficients of determination ( R 2 ), and per capita migration rate m (Dvison et al. 1986). 2.6 Statistical Analyses Our statistical analyses were performed in R 4.4.2 (R Core Team 2024), using R packages: “ vegan ” and “ spaa ” package for VPA and RDA analysis (Zhang 2016; Oksanen et al. 2020); “ Hmisc ”, “ minpack.lm ” and “ stats4 ” package for NCM analysis (Elzhov et al. 2023; Harrell 2023);“ boot ” package for bootstrapping analysis (Davison & Hinkley 1997; Angelo & Ripley 2024). 3 Results 3.1 Diversity and Composition of Mammal Communities From 2019 to 2022, data from 160 cameras, totaling 44,393 camera-trap days, captured an estimated 15,507 independent occurrence records from 21 mammal species (Table S1). Moreover, Amur tiger was distributed in both Xiaobeihu and Hunchun forests, while Amur leopard was recorded Hunchun but not Xiaobeihu forests (Table S2). Top-predator (Amur tiger, Amur leopard, black and brown bear) detected occupancy and relative abundance index (RAI) of Hunchun forest were 2.1× higher than Xiaobeihu ( p <0.05; Table S2). In what follows we refer to Hunchun as top-predator-abundant forest (TPAF), and Xiaobeihu as top-predator-low-abundance forest (TPLF). Beyond differences in top-predator abundance (above), mammals in general were not more abundant in TPAF than in TPLF. Small herbivores (mice, squirrels) were more abundant in TPLF than TPAF, whereas large herbivores and mesopredators did not differ significantly in relative abundance index between TPAF and TPLF. Mammal alpha diversity was not significantly different between TPAF and TPLF ( p =0.071; Figure 1 ; in detail TPAF alpha diversity was slightly higher than TPLF) and neither was Shannon diversity ( p =0.25; Figure 1 ); Shannon diversity was also not significantly different between TPAF-C and TPAF-R ( p =0.14; Figure 1 ). Figure 1 Alpha diversity of mammal community composition and spatial differences (Wilcox test, p values at the top of each panel). a, c, between TPLF (top predator low-abundance forest in Xiaobeihu) and TPAF (top-predator-abundant forest in Hunchun); c, d, Community differences between China and Russia sides (TPAF-C vs. TPAF-R) of Hunchun highway. The box represents 1 st –3 rd quartiles, the central bar represents the median. Mammal community dissimilarity (beta diversity) within TPAF was 1.7× higher than within TPLF locations ( p =0.001; Figure 2a ). In TPAF (Hunchun) locations, mammal beta diversity within the Russia side of Hunchun 331 National Highway (a major highway) was 1.4× higher than within the China side of the highway ( p =0.001; Figure 2b ). In addition, year-to-year variation in beta diversity between mammal metacommunities was not significant ( p >0.05; Table S3). Similarity Percentage (SIMPER) analysis showed a small number of significant differences in mammal metacommunity composition between TPAF and TPLF ( Figure 2c ): badger (10.3% of total community relative abundance) had lower RAI in TPAF than TPLF, and black bear (0.6%) was more abundant in TPAF than TPLF (Table S4). Mammal species also had significant community differences between the China and Russia sides of the Hunchun highway ( Figure 2d ): sika deer (53.3% of total RAI for the community) and hare (1.1%), were more abundant on the Russia side, whereas roe deer (25.6%) was more abundant on the China side. FIGURE 2 Bray-Curtis Rank differences (analysis of similarities, ANOSIM). (a), Spatial metacommunity differences between TPLF (top predator low-abundance forest in Xiaobeihu) and TPAF (top-predator-abundant forest in Hunchun); (b), Community differences between China and Russia sides (TPAF-C vs. TPAF-R) of Hunchun highway. c, d, Species average RAI, Relative Abundance Index (camera trap records per 100 days) and statistical differences measured through similarity percentages (SIMPER) between (c), TPAF and TPLF; (d), TPAF-C and TPAF-Russia. Detailed data are shown in Table S4 and S5. Note: “mice” represents the sum of all nocturnal rodents’ RAI; * p <0.05, ** p <0.01, *** p <0.001. 3.2 Habitat Factors Affecting Mammal Metacommunity Composition Variation partitioning analysis (VPA) showed that across all sites (TPAF and TPLF), geographical location and habitat/environmental factors explained 29.7% of the variance in community composition, with 5.8% assigned to location, 5.8% for habitat/environmental factors and 18.1% shared between habitat/environmental factors and location ( Figure 3a ); the shared variation could indicate that either habitat factors vary spatially or result from dispersal. RDA showed that elevation, geographical factors (longitude, latitude variable is not independence), forest type (broad-leaved mixed, oak) and NDVI were the main factors explaining the differences between TPAF vs. TPLF mammal metacommunity composition ( Figure 3f ; Table S6). In TPLF only 21.6% of the variation in community composition could be explained; habitat/environmental factors explained 10.8%, location explained 4.2%, and 5.5% of variation was shared between habitat/environmental factors and location ( Figure 3c ). For communities in TPAF, much more variation could be explained (52.6%), and the majority was due to location (36.4%) with 2.7% due to habitat/environmental factors and 13.5% shared between habitat/environmental and location factors ( Figure 3b ). Separately analyzing both sides of the Hunchun highway showed that these locational differences in composition for TPAF were because of differences between the two sides of the highway; on the Russia side location explained 3.3%, while the Chinese side it was 8.3% ( Figure 3d, 3e ). RDA showed that elevation, cattle RAI and location (latitude and longitude) were significant factors explaining China vs. Russia side of Hunchun highway differences in mammal community composition ( Figure 3g ; Table S7). FIGURE 3 Analysis of explanatory factors for mammal metacommunity composition. (a-e) Redundancy Analysis, RDA; (f, g) Variance Partitioning Analysis (VPA) explained adj R 2 of different explanatory variables in RDA (f, n=160; g, n=82); In (f, g) red arrows indicate significant variables ( p < 0.05) and gray arrows indicate non-significant variables. TPLF is top predator low-abundance forest; TPAF, is top-predator-abundant forest; NDVI, Normalized Difference Vegetation Index; Mix, mixed coniferous and broad-leaved forest; DBH, average diameter at breast height of trees. The detailed parameters (e.g., units) are shown in Tables S7 and S8, and correlations among explanatory variables in (f, g) are given in Figure S2. 3.3 Geographical Factors Affecting Mammal Community Composition Bootstrapping analyses showed that there was a distance-decay relationship (DDR) in similarity of the TPAF mammal metacommunities. The DDR pattern in both TPAF and TPLF was significant ( p <0.001), but while R 2 for similarity vs. distance was 0.21 for TPAF ( Figure 4b ) it was 0.002 for TPLF ( Figure 4a ). We also found that when we separately considered China vs. Russia sides of the highway (Figure S1), only the DDR on the Russian side of the highway was significant and had a reasonable R 2 -value ( R 2 =0.12; Figure 4d ). This shows that the highway increased the predictability of the DDR. Overall, VPA also explained 30.5% of variation on the Russia side and 14.4% on the China side ( Figure 3d, 3e ); this greater structuring on the Russia side was due to a mix of habitat/environmental factors and interactions of these with location. FIGURE 4 Distance-decay Relationships (DDR) with linear regression analysis (bootstrapping 1000 times) between geographical distance (between pairs of camera-trap locations) and mammal metacommunity composition similarity in top-predator low-abundance forest (TPLF) vs. top-predator-abundant forest (TPAF), and for TAPF China vs. Russia sides (TPAF-C vs. TPAF-R) of Hunchun highway. The mean (±SE) of R 2 and P values were randomly sampled 1000 pairs of camera-trap locations with different geographical distances. 3.4 Mammal Neutral Community Analyses Neutral community models (NCMs) were capable of explaining RAI and occurrence of the mammal communities, with overall R 2 >0.73. The proportion of species that fit NCM predictions of RAI/occurrence values in TPLF vs TPAF varied little ( Figure 5 ; the small number of species precludes a statistical test of the proportion of species). However, there were some interesting differences in RAI and occurrence between sites which are consistent with effects of species interactions. Comparing TPAF sites with all top predators’ present vs sites that lacked the top predators, five of 14 species changed in their NCM-predicted RAI and occurrence values. Not surprisingly given that top predators were at low abundance in TPLF sites, comparisons of TPLF sites with all top predators present and those lacking top predators were associated with only one change of fit of occurrence and RAI relative to the NCM predictions of 13 comparisons. However, comparing TPLF sites with and without a smaller predator, sable, produced five changes in the fit of occurrence and RAI relative to the NCM predictions. Hence, there was good evidence for predators affecting the abundance of other community members in both TPLF and TPAF. In TPAF top predators affected the RAI/occurrence of other species whereas in TPLF, the presence of sable affected an equal number of species. Consistent with predation, a prey species of large predators, Siberian roe deer was above the NCM-predicted RAI and occurrence values (A in Figure 5) in TPLF but was reduced to fitting NCM-predicted values in TPAF (N in Figure 5). Consistent with mesopredator release, badger was below NCM-predicted RAI/occurrence values in TPAF sites with all predators present but fit NCM predictions in TPAF sites lacking top predators and all TPLF sites. Other small predators showed opposite changes in RAI/occurrence: leopard cat, yellow-throated marten, and Siberian weasel were above NCM-predicted RAI/occurrence values (A) in TPAF sites with all predators present but were reduced to fitting NCM-predicted values in TPAF sites lacking all top predators (and some TPLF sites for some of these species); similarly, raccoon dog was above NCM-predicted values in TPAF but fit NCM predictions in TPLF. Small prey species also showed differences in abundance in sites that varied in top-predator abundance: Manchurian hare and Amur hedgehog were above NCM predictions in most TPAF sites but fit NCM predictions in some TPLF sites; similarly Siberian chipmunk and mice fit NCM predictions for RAI/occurrence in most TPAF sites but were below NCM predictions in all TPLF sites (except those lacking sable, a small predator). FIGURE 5 Effects of different top predator species on mammal community assembly between top predator low-abundance forest (TPLF) and abundant forest (TPAF) based on Neutral Community Model (NCM) fitting. Each column shows results of an NCM. The first column in each area shows the complete mammal community, whereas other columns show the model results for locations without the top predator species detected (gray solid blocks); “N” represents that the occurrence frequency of this species was within the prediction of the neutral community model, blue background “A” indicates that the occurrence frequency of this species was above the NCM prediction, red background “B” indicates that the occurrence frequency of this species was below the NCM prediction; gray solid blocks represent that camera-trap locations with this species detected were removed from the NCM shown in that column; the “ R square” is the coefficient of determination for fit of the NCM; “ m value” is the dispersal rate per capita of the NCM; non-black font colors (with underlining) represent state-changing species compared to the model from locations with the maximum number of species present in that area. Note: the low occurrence frequency of red deer and musk deer would have caused their NCM fits to be “N”, and so they are not shown in this Figure. Comparing TPLF locations with vs. without sable (across all three years of the study; n=50), the absence of sable decreased the relative abundance index (RAI) and occurrence frequency of squirrel, chipmunk and mice; these changes also altered the fits of these species to the NCM ( Figure 6a ). Overall, we infer that there is a positive effect of the presence of sable (or the places they inhabit) on squirrel, chipmunk and mice occurrence and RAI. The presence of top predators in TPAF led to reduced occurrence and RAI of marten, but increased the RAI and occurrence of weasel, leopard cat, hedgehog and badger ( Figure 6c ). Overall, more species matched NCM predictions for RAI and occurrence in sites without top predators in TPAF (73%) than sites with top predators (37%; Figure 6c ; p =0.05 in a G-test). Black bears also had substantial effects on the mammal community; we describe the full results for this species because they best illustrate some of the effects of species interactions; effects of leopard, brown bear and tiger in both TPAF and TPLF, and all top predators in TPLF, were less frequent or smaller and are shown in Figure 6b . Black bear was absent in TPLF (locations, n=66), and in these locations the RAI and occurrence frequency of hare was greater and RAI of leopard cat was lower than in sites where black bear was present (producing changes from N to A, above the NCM prediction, and A to N, respectively; compare columns 1 vs. 3 in Figure 5 ; Figure 6b ). Black bears were more abundant in TPAF than TPLF (Table S2) and were above the NCM predicted RAI for their occurrence frequency ( Figure 5 ). In TPAF locations where black bears were not detected, the RAI of chipmunks, weasels, hedgehogs and leopards decreased compared to when black bears were present (from N to A; columns 6 vs. 10 in Figure 5 ; Figure 6d ). Slightly more species matched NCM predicted occurrence/RAI values when black bear was absent (50%) than when they were present (37%) and this effect was only seen in TPAF sites (similar values were 50% vs. 47% in TPLF). Hence, black bears affected the RAI or occurrence frequency of up to four different species and effects were greater when black bears were at high abundance (in TPAF cf. TPLF). Apparent differences between in estimates for per capita dispersal ( m ) and R 2 between TPAF and TPLF, or on different sides (0.722) of the highway in TPAF, or with vs. without particular predators (Figure 5), were not significant (all p >0.05 based on bootstrapped estimates; Table S8). FIGURE 6 Predictions of frequency of occurrence and mean relative abundance compared to the Neutral Community Model (NCM) process of mammal community with changes after removing the locations of the top predator species in top-predator-abundant forest (TPAF) and low-abundance forest (TPLF). The yellow arrows indicate the changing process of each species. The pie chart represents the proportion of species richness prediction below predication (orange), above predication (blue), and not different from neutral (gray). R 2 is the coefficient of determination for fit of the NCM; m is the dispersal rate per capita of the NCM; See supplementary information Figure S3 for detailed NCM species changes. Several mammal species also showed differences in RAI on the China and Russia sides of the major highway in TPAF ( Figure 2 ), which led us to conduct separate NCM analyses for each side of the highway ( Figure 5 ). Doing this showed that while overall in TPAF the occurrence frequency of sika deer was below the NCM prediction, when we separately considered the China and Russia sides of the highway, the occurrence frequency became above predicted (A) on both sides of highway ( Figure 5 ). This likely indicates that low dispersal across the highway reduced sika deer occurrence on the China side; per-capita movement rate estimates for all species from the NCM’s were m= 0.72 on the Russia side compared to only m= 0.40 for China side ( Figure 5 ). 4 Discussion Comparison of large forest areas with relatively static presence and abundance of top predators enabled us to test the role of top predators in the assembly of mammal metacommunities. Our most major finding is that mammal metacommunity assembly was not a zero-sum process. Variation partitioning analysis (VPA) and redundancy analysis (RDA) indicated substantial roles of habitat/environmental structuring of community composition, and substantial roles of habitat/environmental factors interacting with geographical location ( Figure 3 ). These effects were strongest in forest areas with abundant top predators of several species ( Figure 3 ), and beta diversity was also greatest in these areas ( Figure 2 ). Top predators also had substantial effects on the relative abundance and frequency of occurrence of other mammal species, indicating the effects of species interactions ( Figures 5, 6 ). In areas with few top predators (TPLF), other species were affected by a small predator species, sable, rather than top-predator species (Figure 5). Moreover, the large block of forest with the most top predators (Hunchun or TPAF, Russia side of the highway) also showed the strongest distance decay relationships of community composition ( Figure 4 ). Hence, strong effects of habitat/environmental structuring were coincident with strong effects of species interactions, and high estimates of community-wide dispersal! While techniques such as VPA have been shown to have many limitations (e.g. (Brown et al. 2017), it is clear that the fundamental assumption of assembly processes having a zero sum is not valid. Our analyses provided no indication of a tradeoff with strong habitat filtering being accompanied by weakened effects of species interactions or dispersal in structuring (meta)communities. While abundant top predators altered beta diversity and the relative abundance and frequency of occurrence of other mammal species, large forest areas with a long-term absence or low abundance of top predators showed effects of mesopredators were involved in species interactions. While mesopredator release is a well-studied phenomenon (Saggiomo et al. 2021), intuitively we might expect based on well-studied effects of top predators (Soulé et al. 2005; She et al. 2025) that top predators would have larger effects on metacommunity assembly processes. Instead, we have an equal amount of evidence that mesopredators altered the relative abundance and frequency of occurrence across sites of other mammal species when top predators were absent or at low abundance ( Figures 5, 6 ). In general, forests with low abundance of top predators also had less habitat/environmental structuring as revealed by VPA and better fits (higher R 2 ) of NCMs ( Figures 5, 6 ). 4.1 Stochastic Processes Mainly Affect Mammal Metacommunity Assembly Notwithstanding the VPA and RDA analyses (Figure 3), which showed substantial variation in community composition was accounted for by habitat/environmental or geographical location (or their interaction), NCM analyses indicated that mammal community composition was mainly dominated by stochastic processes (NCM R 2 >0.73; Figure 5 ). This is surprising given that species interactions and habitat selection are often mainly assumed to affect deterministic processes (Vellend & Agrawal 2010). A possible explanation is that NCM’s are highly flexible and the low amount of information in static occurrence and abundance (RAI) data mean that NCM’s could fit these data well. The habitat/environmental factors in VPA and RDA, and the effects of the presence of different predator species on the occurrence and RAI provide substantially more information. We therefore do not regard the different types of analysis as being contradictory. High beta diversity and strong DDR in TPAF compared to TPLF suggest that communities are also structured by deterministic habitat factors and/or species interactions ( Figure 2, 4 ). Only the Russia side of the highway had substantial dispersal limitation or habitat structure over distance ( Figure 4 ) shown by community similarity decreasing with increasing geographical distance. We also found that distance from villages and the highway were important drivers of the composition of Hunchun mammal communities, which is consistent with (Miquelle et al. 2015). 4.2 Predator-prey interactions and reduced stochastic processes in mammal community assembly In TPLF, we found that badger and rodents (squirrel, chipmunk and mice) increased their NCM predicted RAI/occurrence values (N to A or B to N) comparing habitats without sable to locations with these species ( Figures 6a ). Based on known predation relationships between sable and rodents (chipmunks and mice), we believe this is likely caused by mesopredator release of sable in TPLF (Chutipong et al. 2017). Strictly speaking, because we compared sites with and without sable, this could either be because of differences in predator-prey interactions or habitat-associated factors. Even in locations where top predators were not detected, we did not find that the occurrence frequency of rodents (i.e., squirrel, chipmunk and mice) was below NCM-predicted values ( Figure 2 ). We believe that the top-down effect of top predators was replaced by mesopredator release in areas of the forest that lacked top predators (She et al. 2023). We also found that hare and chipmunk changed from neutral processes to above predicted occurrence/RAI (N-A) in locations with vs. without black bear (Figure 6b and 6d). This could be because of black bears as predators but it has also been observed that some rodents and bears have correlated abundances and co-occurrence after emergence from hibernation (Enders & Vander Wall 2012). Our study confirms the role of top predators (e.g. Amur leopard) in modulating forest generalist species (e.g., badgers), which could contribute to vegetation restoration (Xu et al. 2023); although wild boar are also generalist species and their NCMs status was not significantly affected by top predators, their nesting can also be beneficial to reshape the composition of vegetation communities (Luskin et al. 2021); such effects illustrate the importance of top predators to forest conservation. In conclusion, our study found that mammal communities in large areas of forest that are relatively static in having high abundance of top predators showed strong effects on community composition of species interactions and habitat filtering and some effects of dispersal on community composition. We caution against assuming that community assembly processes are zero-sum and instead suggest that habitat/environmental structuring may set high abundances and diversity, setting the stage for strong species interactions and structured communities over space despite high dispersal rates. We also emphasize the value of applying multiple analytical methods at multiple scales and using large-scale variation in species distributions to understand assembly processes. Acknowledgments This research was funded by the National Key Research and Development Program of China: Migration and diffusion mechanism of wild animals and population control technology (2023YFF1305000). Statement of authorship GJ conceived and design this research. BH, NJR, DL, WC, DW, JW and YG performed field working and collected data. B.H. analyzed output data. B.H. wrote the first draft of the manuscript, and NJR., HB, JQ, MH and GJ contributed substantially to revisions. Competing interests: The authors declare no competing interests. References www.webmap.cn http://www.gscloud.cn Angelo, C. & Ripley, B. (2024). boot: Bootstrap R (S-Plus) Functions .Biswas, S.R. & Wagner, H.H. (2012). Landscape contrast: a solution to hidden assumptions in the metacommunity concept? Landscape Ecology , 27, 621-631.Borcard, D., Legendre, P. & Drapeau, P. (1992). Partialling out the Spatial Component of Ecological Variation. Ecology , 73, 1045-1055.Brown, B.L., Sokol, E.R., Skelton, J. & Tornwall, B. (2017). Making sense of metacommunities: dispelling the mythology of a metacommunity typology. Oecologia , 183, 643-652.Burns, A.R., Stephens, W.Z., Stagaman, K., Wong, S., Rawls, J.F., Guillemin, K. et al. (2016). Contribution of neutral processes to the assembly of gut microbial communities in the zebrafish over host development. Isme Journal , 10, 655-664.Cao, H., Hua, Y., Liang, X., Long, Z., Qi, J., Wen, D. et al. (2022). Wavelet Analysis Reveals Phenology Mismatch between Leaf Phenology of Temperate Forest Plants and the Siberian Roe Deer Molting under Global Warming. Remote Sensing , 14, 3901.Chase, J.M. & Myers, J.A. (2011). Disentangling the importance of ecological niches from stochastic processes across scales. Philosophical Transactions of the Royal Society of London , 366, 2351-2363.Chen, B., Peng, Z., Chen, S., Liu, Y., Qi, J., Pan, H. et al. (2024). Bridging ecological processes to diversity formation and functional profiles in belowground bacterial communities. Soil Biology & Biochemistry , 198, 109573.Cheng, M., Luo, S., Zhang, P., Xiong, G., Chen, K., Jiang, C. et al. (2024a). A genome and gene catalog of the aquatic microbiomes of the Tibetan Plateau. Nature Communications , 15, 1438.Cheng, W., Gray, T.N.E., Bao, H., Wen, D., Liang, X., She, W. et al. (2024b). Drivers of Human–Tiger Conflict Risk and Potential Mitigation Approaches. Ecosphere , 15, e4922.Chutipong, W., Steinmetz, R., Savini, T. & Gale, G.A. (2017). Assessing resource and predator effects on habitat use of tropical small carnivores. Mammal Research , 62, 21-36.Clarke, K.R. (1993). Non-parametric multivariate analyses of changes in community structure. Austral Ecology , 18, 117-143.CNCSGI (2025). Chinese National Catalogue Service for Geographic Information. Available at: .Cottenie, K. (2005). Integrating environmental and spatial processes in ecological community dynamics. Ecology Letters , 8, 1175-1182.Davison, A.C. & Hinkley, D.V. (1997). Bootstrap Methods and Their Applications . Cambridge University Press, Cambridge.Dickman, A., Cotterill, A., Asecheka, S., Mlaponi, Z., Mbugi, H., Grau, A. et al. (2025). Community Camera Trapping: A Novel Method for Encouraging Human–Big Cat Coexistence on Human-Dominated Land. Wildlife Letters , 3, 22-29.Dvison, A.C., Hinkley, D.V. & Schechtman, E. (1986). Efficient bootstrap simulation. Biometrika , 73, 555-566.Elzhov, T., Mullen KM, Spiess A & B, B. (2023). minpack.lm: R Interface to the Levenberg-Marquardt Nonlinear Least-Squares Algorithm Found in MINPACK, Plus Support for Bounds .Enders, M.S. & Vander Wall, S.B. (2012). Black bears Ursus americanus are effective seed dispersers, with a little help from their friends. Oikos , 121, 589-596.GDCs (2025). Geospatial Data Cloud site. Available at: .Harrell, J.F. (2023). Hmisc: Harrell Miscellaneous .Holyoak, M., Caspi, T. & Redosh, L.W. (2020). Integrating Disturbance, Seasonality, Multi-Year Temporal Dynamics, and Dormancy Into the Dynamics and Conservation of Metacommunities. Frontiers in Ecology and Evolution , 8, 571130.Jiang, G., Wang, G., Holyoak, M., Yu, Q., Jia, X., Guan, Y. et al. (2017). Land sharing and land sparing reveal social and ecological synergy in big cat conservation. Biological Conservation , 211, 142-149.Jiao, S., Chu, H., Zhang, B., Wei, X., Chen, W. & Wei, G. (2022). Linking soil fungi to bacterial community assembly in arid ecosystems. iMeta , 1, e2.Jin, Y., Liang, Y., Yao, M., Kong, W., Zhu, S., Roberts, N.J. et al. (2025). Identifying Human–Tiger Conflict Risk and Priority Management Areas in Laoyeling, Northeast China. Wildlife Letters , 3, 7-15.Johnson, C.N. & Fisher, I.D.O. (2007). Rarity of a top predator triggers continent-wide collapse of mammal prey: dingoes and marsupials in Australia. The Royal Society Proceedings B , 274, 341-346.Legendre, P.L. & Legendre, L.F.J. (2012). Numerical Ecology . Elsevier.Leibold, M.A., Holyoak, M., Mouquet, N., Amarasekare, P., Chase, J.M., Hoopes, M.F. et al. (2004). The Metacommunity Concept: A Framework for Multi-Scale Community Ecology. Ecology Letters , 7, 601-613.Letnic, M., Koch, F., Gordon, C., Crowther, M.S. & Dickman, C.R. (2009). Keystone effects of an alien top-predator stem extinctions of native mammals. Proceedings of the Royal Society B: Biological Sciences , 276, 3249-3256.Li, Z., Wang, H., Ge, J. & Wang, T. (2024). Spatiotemporal patterns of small carnivores in a human-dominated forest landscape shared with apex predators. Landscape Ecology , 39, 214.Liu, J., Kang Y, Jensen AJ, Kays R & A., J. (2025). Apex predator loss drives trophic downgrading in China’s protected areas. Current Biology , 35, 2872-2880.Luskin, M., Johnson, D., Ickes, K., Yao, T. & Davies, S. (2021). Wildlife disturbances as a source of conspecific negative density-dependent mortality in tropical trees. Proceedings. Biological sciences , 288, 20210001.Marrec, R., Pontbriand‐Paré, O., Legault, S. & James, P.M.A. (2018). Spatiotemporal variation in drivers of parasitoid metacommunity structure in continuous forest landscapes. Ecosphere , 9, e02075.Minchin, P.R. (1987). An evaluation of the relative robustness of techniques for ecological ordination. Vegetatio , 69, 89-107.Miquelle, D.G., Rozhnov, V.V., Ermoshin, V., Murzin, A.A. & Naidenko, S.V. (2015). Identifying ecological corridors for Amur tigers ( Panthera tigris altaica ) and Amur leopards ( Panthera pardus orientalis ). Integrative Zoology , 10, 389-402.O’Brien, T.G., Kinnaird, M.F. & Wibisono, H.T. (2003). Crouching tigers, hidden prey: Sumatran tiger and prey populations in a tropical forest landscape. Animal Conservation , 6, 131-139.O’Connell, M.A. & Hallett, J. (2019). Community ecology of mammals: deserts, islands, and anthropogenic impacts. Journal of Mammalogy , 1019-1043.Oksanen, J., Blanchet FG, Friendly M, Kindt R, Legendre P, McGlinn D et al. (2020). vegan: Community Ecology Package .R Core Team (2024). R: A Language and Environment for Statistical Computing , Vienna, Austria.Riddley, M., Hepp, S., Hardeep, F., Nayak, A., Liu, M., Xing, X. et al. (2025). Differential roles of deterministic and stochastic processes in structuring soil bacterial ecotypes across terrestrial ecosystems. Nature Communications , 16, 2337.Saggiomo, L., Bar, V. & Esattore, B. (2021). The fox who cried wolf: A keywords and literature trend analysis on the phenomenon of mesopredator release. Ecological Complexity , 48, 100963.Selva, N., Hobson, K.A., Zalewski, A., Cortes-Avizanda, A. & Donázar, J.A. (2024). Mammal communities of primeval forests as sentinels of global change. Glob. Change Biol. , 30, e17045.She, W., Gu J, Qi J, Liu S, Yang X, Zhang Z et al. (2024). External and internal driving forces of community and functional group stability in the forest mammal community. Proc. R. Soc. B , 291, 20240775.She, W., Gu, J.Y., Holyoak, M., Yan, C., Qi, J.Z., Wan, X.R. et al. (2023). Impacts of top predators and humans on the mammal communities of recovering temperate forest regions. Sci. Total Environ. , 862, 160812.She, W., Holyoak, M., Gu, J., Qi, J., Liu, S. & Jiang, G. (2025). Abundant top predators increase species interaction network complexity in northeastern Chinese forests. J. Anim. Ecol. , 94, 745-759.Sloan, W.T., Lunn, M., Woodcock, S., Head, I.M., Nee, S. & Curtis, T.P. (2006). Quantifying the roles of immigration and chance in shaping prokaryote community structure. Environmental Microbiology , 8, 732-740.Soulé, M.E., James A. Estes, Brian Miller & Douglas L. Honnold (2005). Strongly Interacting Species: Conservation Policy, Management, and Ethics. BioScience , 55, 168-176.Spiesman, B.J. & Inouye, B.D. (2013). Habitat loss alters the architecture of plant–pollinator interaction networks. Ecology , 94, 2688-2696.Sytsma, M.L.T., Lewis, T., Bakker, J.D. & Prugh, L.R. (2023). Successional patterns of terrestrial wildlife following deglaciation. J. Anim. Ecol. , 92, 723-737.Thompson, P.L. & Gonzalez, A. (2017). Dispersal governs the reorganization of ecological networks under environmental change. Nat. Ecol. Evol. , 1, 0162.Tylianakis, J.M., Didham, R.K., Bascompte, J. & Wardle, D.A. (2008). Global change and species interactions in terrestrial ecosystems. Ecology Letters , 11, 1351-1363.Vellend, M. & Agrawal, A. (2010). Conceptual Synthesis in Community Ecology. The Quarterly Review of Biology , 85, 183-206.Wang, Y., Cheng, W., Guan, Y., Qi, J., Roberts, N.J., Wen, D. et al. (2023). The fine‐scale movement pattern of Amur tiger ( Panthera tigris altaica ) responds to winter habitat permeability. Wildlife Letters , 1-12.Warton, D.I., Wright, S.T. & Wang, Y. (2012). Distance‐based multivariate analyses confound location and dispersion effects. Methods in Ecology & Evolution , 3, 89-101.Wen, D.S., Qi, J.Z., Cheng, W.N., Li, Z.Y., Qi, Q., Cui, Y.L. et al. (2024). Spatial population distribution dynamics of big cats and ungulates with seasonal and disturbance changes in temperate natural forest. Glob. Ecol. Conserv. , 51, e02881.Xiao, Z., Chen, L., Song, X., Shu, Z., Xiao, R. & Huang, X. (2019). Species inventory and assessment of large- and medium-size mammals and pheasants using camera trapping in the Chebaling National Nature Reserve, Guangdong Province Biodiversity Science , 27, 237-242.Xu, C., Silliman, B.R., Chen, J., Li, X., Thomsen, M.S., Zhang, Q. et al. (2023). Herbivory limits success of vegetation restoration globally. Science , 382, 589-594.Yang, H., Han, S., Xie, B., Mou, P., Kou, X., Wang, T. et al. (2019). Do prey availability, human disturbance and habitat structure drive the daily activity patterns of Amur tigers ( Panthera tigris altaica )? Journal of Zoology , 307, 131-140.Zhang, J. (2016). spaa: Species Association Analysis .Zhou, Y.Y. & Wang, J.P. (2023). The Composition and Assembly of Soil Microbial Communities Differ across Vegetation Cover Types of Urban Green Spaces. Sustainability , 15, 13105. Information & Authors Information Version history V1 Version 1 22 September 2025 Copyright This work is licensed under a Non Exclusive No Reuse License. Keywords amur tiger community assembly metacommunity neutral community model (ncm) top predators Authors Affiliations Baoxiang Huang 0000-0003-3076-3239 Northeast Forestry University View all articles by this author Marcel Holyoak 0000-0001-9727-3627 University of California View all articles by this author Nathan Roberts 0000-0001-8299-4499 Northeast Forestry University View all articles by this author Dongqi Liu Northeast Forestry University View all articles by this author Wannian Cheng Northeast Forestry University View all articles by this author Dusu Wen Southwest Forestry University View all articles by this author Jingxuan Wang Chinese Academy of Sciences View all articles by this author Yanhui Guan Heilongjiang Xiaobeihu National Nature Reserve Administration View all articles by this author Heng Bao Northeast Forestry University View all articles by this author Jinzhe Qi Northeast Forestry University View all articles by this author Guangshun Jiang 0000-0001-6321-9489 [email protected] Northeast Forestry University View all articles by this author Metrics & Citations Metrics Article Usage 234 views 147 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Baoxiang Huang, Marcel Holyoak, Nathan Roberts, et al. Assembly Processes Structuring Mammal Metacommunities are not Zero-sum. Authorea . 22 September 2025. DOI: https://doi.org/10.22541/au.175856347.70125992/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.175856347.70125992/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9ffc43afda32e2c5',t:'MTc3OTQ1NzU1OQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();