Assessing abiotic correlations of an indicator species with sympatric riparian birds in a threatened submontane river–forest system using joint species modelling

To assess the abiotic correlations between indicator species and others in the community. Additionally, we evaluated the importance of environmental variables in driving the co‐occurrence patterns of the coexisting riparian bird species within a threatened aquatic and terrestrial system.


| INTRODUC TI ON
The complexity and severity of problems related to environmental degradation and biodiversity loss have promoted the development of indicator species and other surrogate approaches to track changes in environments and biodiversity (Canterbury et al., 2000). Using indicator species or species groups to represent the environmental needs of a wider range of similar species contributes to simplifying conservation, management planning, and communication (Siddig et al., 2016). Identifying indicator species often relies on knowledge of the species' "realized niche", where species abundance, presence, or use data are used to identify important abiotic and biotic factors driving the species' distribution (Poggiato et al., 2021). Nevertheless, few studies have thus far been considered to distinguish the pure abiotic environmental effects (i.e., fundamental niche; Poggiato et al., 2021) from the effects of species interactions in the process of identifying indicator species. The limited biological knowledge of how indicator species associate with target environmental characteristics or taxa imposes a unique challenge to their application in conservation actions.
Joint species distribution models (JSDMs) provide a framework that considers both environmental filtering and biotic interactions in the identification of indicator species by simultaneously quantifying the effects of the abiotic environment and introducing latent variables to explain variation in species composition not accounted for by measured predictor variables (e.g., potential biotic interactions; Pollock et al., 2014;Warton et al., 2015). Note that latent variables are used only to infer the biotic signals of potential species associations rather than strictly providing evidence for proven species interactions (Dormann et al., 2018). Despite their conceptual advantages, the application of latent variables to identify the fundamental mechanisms behind species coexistence is still controversial based on correlational evidence (Blanchet et al., 2020;Poggiato et al., 2021). However, in contrast to other methods exploring species distribution and co-occurrence patterns, such as species distribution models, canonical correspondence analysis (ter Braak, 1986) and pairwise co-occurrence analysis (Veech, 2013), the JSDM has exclusive advantages: an ability to partition the effect of measured predictors and residual correlations between species, and predictions of community assemblages that account for these correlations.
By partitioning the effect of measured variables from residual correlations, JSDMs potentially enable more accurate estimates of environmental drivers of species distributions. In addition, by leveraging the information in the residual correlations, JSDMs can potentially better predict community assemblages (Wilkinson et al., 2020).
Riparian zones are recognized as focal areas for biodiversity due to their special spatial positions (interface of water-land), unique microclimates, and other biophysical features and resources (Olson et al., 2007;Palmer & Bennett, 2006). Under anthropogenic disturbances (e.g., hydroelectric generation, farming, and deforestation), riparian habitats have undergone widespread impacts, including vegetation clearing and modification, stream bank erosion, alteration of belowground processes, and reduced water flows and quality, resulting in decreased biodiversity of diverse taxa (Jansen & Robertson, 2001;Kauffman et al., 2004). Nevertheless, over the last few decades, submontane riparian zones have remained among the least inventoried, monitored and conserved sections of rivers compared to riverine habitats in lowlands and floodplains due to their sparse distribution, relatively small extent and limited accessibility (Kajtoch et al., 2016). As a large group of species inhabiting submontane riparian zones, birds are sensitive to ecological conditions and their diversity and community structure are often used as a measure of conservation status or need for restoration by many conservation scientists, land managers, and policy makers (Canterbury et al., 2000;Fleishman et al., 2003). In addition, various interspecific interactions between bird species in mutualistic (e.g., sharing of limited resources) and antagonistic (competition, predation and parasitism) relationships can also signal the condition of an ecosystem (Davis et al., 1998;Gilman et al., 2010). Therefore, quantifying correlations in bird species abundance due to shared environmental responses and their respective residual correlations (responses to unknown covariates or biotic interactions) could extend beyond simply explaining the distribution and coexistence of bird species to advancing the effective development of conservation management plans for bird communities and riparian environments that maintain productivity and ecosystem integrity.
The global endangered scaly-sided Merganser Mergus squamatus is a submontane riparian specialist that feeds in the river and nests in holes of old-growth trees (Kajtoch et al., 2016). This species is considered a potentially good indicator in submontane riparian ecosystems due to its specific preference for habitat fragments with high ecological value (e.g., higher habitat complexity and availability) and a positive correlation with other riparian taxa (Xu et al., 2021).
However, the lack of biological knowledge on co-occurrence relationships with the scaly-sided Merganser and other riparian bird species limited our confidence in the indicator role of this species and the forces driving the composition of riparian bird communities.
Thus, the purpose of this study on bird communities in submontane riparian habitats was to apply JSDMs to assess abiotic correlations of the scaly-sided Merganser with other riparian birds in the species associations in modelling species distributions and to better realizing the roles of indicators in practical applications.

K E Y W O R D S
biological interactions, indicator species, joint species distribution models, riparian habitat, scaly-sided Merganser community, and to evaluate the importance of environmental variables in driving the co-occurrence patterns of the coexisting riparian bird species within the threatened submontane forest-river system.

| Study area
This study was conducted in the submontane valley region of Changbai Mountain in eastern Jilin Province, China (41°51′-44°3′N, 127°32′-129°0′E; Figure 1). Seven river valleys were investigated: the Zhuerduo River, Fuer River, Lushui River, Toudaobai River, Songjiang River, Manjiang River, and Damalu River. All studied river sections are typical submontane riparian areas and are characterized by naturally braided channels with many gravel alluvia, clay scarps and forested banks. Due to the temperate continental monsoon climate, the annual precipitation is approximately 680 mm/year, of which approximately 80% occurs from May to September. July is the hottest month, with an average temperature of 22.8°C, and the coldest is January, at 17.3°C below zero (Sun, 2022). We chose a series of 98 sampling sites based on a stratified random sampling strategy with an elevation range from 351 to 942 meters above sea level. We randomly selected 8-12 sampling sites in each river based on the river length. A distance of more than 1 km was maintained between adjacent sampling sites to prevent spatial autocorrelations (Dorman, 2007).

| Bird survey
We used the point-count method to count birds in May and June 2021 (Hutto et al., 1986), during the peak of the bird breeding season (Padmanabhan & Yom-Tov, 2000). Each sampling site was visited twice by two experienced ornithologists standing at the land-water interface and recording all birds seen or heard in 10 min within a 50-m radius, except for birds that flew through the point-count area. As an estimate of the relative abundance of each bird species at each site, we used the highest value attained from two counts (Toms et al., 2006). Bird surveys were conducted only on clear days F I G U R E 1 Locations of study submontane valleys and sample sites. Upper left inset locates Jilin Province in the Northeast China. Upper right inset shows specific locations of the seven study submontane valleys. Grey dots shows distribution of bird sampling points. within 4 h after sunrise without strong wind and rain, conditions that interfere with the detection and audibility of vocalizations.

| Environmental data
We gathered data on habitat structure at all 98 sites on the same days as the bird counts (Table 1). Three quadrats with 5-m sides were randomly selected within a 50 m radius at each sampling site. For each quadrat, we recorded the height of trees, shrubs, and herbs with a laser rangefinder (Nikon Laser-800) and tapeline. Then, we calculated the vegetation height variability (H SD ) by calculating the standard deviation of the heights of the three types of vegetation (Bae et al., 2018) to reflect vertical heterogeneity in vegetation structure. In addition, we separately recorded the height and DBH (Diameter at Breast Height) of larger trees (DBH ≥ 35 cm) that were retained from historic deforestation. Old-growth forest patches such as these sites have been demonstrated to be positively related to bird richness in many previous studies (Mitchell et al., 2009;Virkkala, 2006). Finally, three quadrat averaged values for each sampling point were included in the subsequent analysis.
Furthermore, the proportion of gravel bars, channel width and sinuosity (the ratio of the distance between two points along the channel and straight line distance between these points) were determined within a 200 × 200 m rectangle around each sampling site through visual interpretation from Landsat satellite imagery. We calculated the Shannon diversity of land cover types to represent habitat complexity according to a 30-m resolution raster data array from www.geoda ta.cn, in which land cover types were divided into 20 categories (such as forest, cropland, water bodies, etc.; see Table S1).
Details of all environmental predictors included in the analyses are shown in Table 1.

| Model construction
We used the "boral" package (Bayesian ordination and regression models for analysing multivariate data in ecology; Hui, 2016) in R v3.6.3 (R Development Core Team, 2019) to quantify the relative importance of the local abiotic environmental conditions and unexplained residuals (i.e., latent variables) of bird communities for each site, and excluded random site effects in our model process based on bird relative abundance (the numbers of individuals of each species that were detected by seen or heard at each site) data. In this study, we first fitted a pure latent variable model (LVM) with two latent variables in which species distribution data were regressed against latent variables, which could be regarded as a model-based unconstrained ordination (Hui, 2016). We then fitted a correlated response model (CRM) to explore the relationships between the bird community and habitat attributes and separate species correlations caused by environmental responses and unknown variables such as interspecific interactions. The MCMC was running with default settings: burn-in 10,000 iterations, total number of iterations including burn-in = 40,000, thinning rate = 30, and normally distributed priors with a mean of zero and a variance of 100 (Hui, 2016). A negative binomial error distribution was used to account for overdispersion in modelling species abundance simultaneously (Haak et al., 2020).
Model convergence was checked via "boral" diagnostic tools using Dunn-Smyth residuals and a normal quantile plot of residuals ( Figures S1 and S2). After fitting, the 95% highest posterior density (HPD) intervals that did not contain zero revealed significant correlations between species relative abundance and environmental predictors. Finally, we calculated the variance in the trace of the estimated residual covariate matrix between the LVM and CRM to obtain an overall measure of the model's predictive power (Warton et al., 2015).

| RE SULTS
We collected a total of 1532 records of 77 species across the 98 surveyed points. To avoid convergence problems, we removed rare species, defined as those present at fewer than five sites from the dataset (Planillo et al., 2020). Ultimately, 1316 records of 26 bird species remained in the bird relative abundance matrix (Table S2) The relative importance of environmental predictors for each bird species was identified by using the 95% HPD (Figure 3). We found that the relative abundance of birds was affected by landscape and local environmental variables simultaneously. Among all abiotic covariates, land cover diversity, proportion of gravel bars, and channel width affected the largest number of bird species. In addition, we found that most bird species had a nonsignificant response to the old-growth tree DBH, but the proportion of these trees served as an exceptional predictor of the relative abundance of scaly-sided Merganser and Mandarin duck.

| DISCUSS ION
This study examined the community structure of riparian birds, including aquatic and terrestrial taxa, while accounting for the effects of environmental and potential species associations by using JSDM, with a focus on abiotic correlations of the scaly-sided Merganser with other coexisting birds in a threatened submontane river-forest ecosystem.
In the control of residual correlations, co-occurrence patterns were dominantly driven by numerous positive correlations. The results are necessary for verifying the role of scaly-sided Merganser as an indicator species. This study provides a robust method for justifying the use of indicator species, which in most cases is overlooked in conservation (McGeoch et al., 2002). Correlations due to abiotic covariates reveal that the coexistence and replacement between studied riparian birds develops based on their similar or opposite niche requirements.
Our results showed that the scaly-sided Merganser could indeed act as a reliable biodiversity indicator in the submontane riparian zones of the Changbai Mountains, considering that more than 70% of the riparian bird species were associated with the scaly-sided Merganser.
Compared with a previous study (Xu et al., 2021), we confirmed that this strong correlation of co-occurrence was primarily driven by abiotic factors, suggesting that riparian habitats with scaly-sided Merganser occurrence may provide niches related to higher productivity, heterogeneity, and food availability for other birds. These bird species seemed to respond to environmental covariates in a parallel way, so they likely should be sensitive to environmental changes in a similar way (Han et al., 2020) and might jointly benefit from protection plans focused on the recovery and maintenance of riparian habitats. F I G U R E 2 Species correlations due to the shared environmental responses (a) and residual covariates (b). Only significant correlations, based on 95% credible intervals excluding zero are shown. The colour of circles (blue and red) represents positive and negative correlations, respectively, and the size of circles represent the strength of the correlation.
Species distributions and co-occurrences are jointly affected by a variety of complex ecological processes, such as environmental filtering and biotic interactions (Tulloch et al., 2018). The co-occurrence of riparian bird species cannot be fully explained without taking potential interspecific associations into account (Pollock et al., 2014). Similar to some other studies ( co-occurrence (Monkkonen et al., 1990;Xu et al., 2021). In addition, the potential species associates with other riparian bird species may be attributed to positive and negative biotic interactions. For example, the Mandarin duck may utilize the occurrence of scaly-sided Merganser, which arrives earlier at the breeding sites, as a cue to select higher quality nests. The barn swallow Hirundo rustica, redrumped swallow and tree sparrow Passer montanus could benefit from reduced predation risk and increased foraging efficiency in mixed-species flocks (Fitzgibbon, 1990;Pulliam, 1973). The competition for food resources in the breeding season may have resulted in the negative associations between yellow-throated bunting and black-faced bunting due to their similar ecological niches. However, several interspecific correlations between bird species as detected by our model seem unlikely to be explained by biological interactions, which may be related to migration, phylogenetic history or unmeasured environmental predictors (Dormann et al., 2018;Zurell et al., 2018), especially for specialists that occupy a unique niche space (Dormann et al., 2018;Han et al., 2020). For instance, the positive residual correlations between the scaly-sided Merganser and Eastern crowned warbler Phylloscopus coronatus indicated that the environmental variables used to model species distributions did not fully capture the variability in habitat. In consideration of the fundamental limitation of the model-based approaches that residual correlations are not equivalent to mechanistic inference of interaction effects (Blanchet et al., 2020;Münkemüller et al., 2020), the pairwise associations we infer here represent a conservative estimation. Nevertheless, separating residual correlations could enable more accurate estimates of the environmental drivers of species coexistence.
The species-specific response of submontane riparian birds to environmental variables showed a dominant effect of land cover diversity, proportion of gravel bars, and channel width on bird relative abundance, which have been identified as focal environmental filters in previous studies on other riparian bird species (Hillman et al., 2016;Sinha et al., 2022;Zeng et al., 2015). Consistent with previous studies (Bae et al., 2018;Cubley et al., 2020), habitat complexity emerged again as the main driver of differences in bird abundance. In this case, river valley fragments containing multiple habitat types could meet the various niche requirements of riparian species Rockwell & Stephens, 2018). First, a more complex riverine landscape composition, which was primarily correlated with bare land and cropland, had a higher number of bird species, involving not only habitat specialists (such as brown dipper Cinclus pallasii) but also a high relative abundance of generalists (e.g., Azure-winged magpie Cyanopica cyanus and red-rumped swallow Cecropis daurica), which may be more adaptive to habitat disturbance. In addition, river fluvial geomorphology (i.e., gravel bars and channel width) significantly affected the relative abundance of aquatic birds, such as common sandpiper Actitis hypoleucos, probably because wider channels could provide aquatic birds greater food availability and refuge space, and gravel bars could alternatively offer habitats with wetting and drying due to changing water levels (Garcia et al., 2012;Jin & Qin, 2020). This study confirmed the importance of environmental variables in driving the co-occurrence patterns of submontane riparian bird communities. Strong positive co-occurrence correlated with environmental factors between most birds suggested that these riparian species may collectively benefit from effective management and conservation actions of riparian zones. Managers should concentrate conservation efforts on river sections with higher habitat heterogeneity, large areas of bars, and wide channels, which may represent areas of high bird biodiversity in submontane riparian habitats (Xu et al., 2021). In addition, the scaly-sided Merganser showed a high frequency of co-occurrence with other riparian birds and significant responses to almost all measured environmental attributes of classic submontane riparian habitats.
The environmental shared responses with other sympatric riparian birds species probably the main reason for this strict physical environment preferences and extensive community connections.
Thus, this species could be used as a valid indicator of high-quality environments and elevated biodiversity in riparian ecosystems.
Accordingly, the management and protection of scaly-sided Merganser could simultaneously benefit the evaluation and monitoring of important habitats more efficiently and help to maintain the biodiversity of other riparian birds, as well as other riparian species. Disentangling the relative importance of abiotic environmental conditions and potential species associations in the roles of indicators could provide insights for conservation biologists and land managers, which contribut to better realizing their indicator roles in practical applications and focusing on limited conservation efforts in more noteworthy aspects (e.g., physical environmental conditions or specific species). Our results emphasize the necessity of incorporating potential species associations in the assessment of the biotic correlations among species in a community and, ultimately, in the identification of the roles of indicator species and in assisting effective management decisions in biodiversity conservation.

ACK N O WLE D G E M ENTS
We thank the two anonymous reviewers for suggestions that greatly improved this manuscript. We thank Zheng Han for his assistance in data analysis. This study was supported by the National Natural Science Foundation of China (No. 32101271 to YG and No. 41901137 to LW), Jilin Province Development and Reform Commission (No. 2021C02), and the Natural Science Foundation of Jilin Province of China (No. 20210101001JC).

CO N FLI C T O F I NTE R E S T S TATE M E NT
We declare that we have no financial and personal relationships with other people or organizations that can inappropriately influence our work, there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in, or the review of, the manuscript entitled "Assessing abiotic correlations of an indicator species with sympatric riparian birds in a threatened submontane river-forest system using joint species modeling".

DATA AVA I L A B I L I T Y S TAT E M E N T
The raw data are available in Appendix S2.