Temporal stability in species richness but reordering in species abundances within avian assemblages of a tropical Andes conservation hot spot

Abstract As the pace of environmental change increases, there is an urgent need for quantitative data revealing the temporal dynamics of local communities in tropical areas. Here, we quantify the stability of avian assemblages in the highly threatened, but poorly studied, Andean biodiversity hot spot. We evaluated the temporal variation in species richness and community composition of local bird assemblages in three habitat types (native forest, introduced forest, native shrub) using a unique, relatively long‐term data series from Cajas National Park and Mazán Reserve in the southern Andes of Ecuador. We sampled birds with mist nets using a standardized protocol over 11 years, from 2006 to 2016. Species richness remained stable over time across habitats, but community composition changed in the native forest. In particular, we observed taxonomic reordering in the native forest, in which the evenness in the distribution of abundances of taxa decreased over time. This finding is consistent with other studies where species richness remained constant over time while community composition changed. Our study highlights the value of long‐term studies in the tropical Andes as we show that species composition of birds in a montane forest is changing, consistent with global trends in biodiversity change.


| INTRODUC TI ON
The increasing pace of environmental change has resulted in a rapid decline of populations, range shifts, and, in some cases, extinctions over the last century (Barnosky et al., 2011;Rosenberg et al., 2019). However, at the local scale species richness, somewhat paradoxically, is often stable over time (Blowes et al., 2019;Dornelas et al., 2014;Hillebrand et al., 2018). Local communities are part of metacommunities (Leibold et al., 2004), and as such, changes in species abundance or extinctions can be compensated for by immigration of individuals from the regional pool (Nielsen et al., 2019). As a result, species richness in local communities might be temporally stable (Magurran & Henderson, 2018), while the composition and abundance of species might be dynamic (Blowes et al., 2019;Larsen et al., 2018). Unfortunately, in many tropical biodiversity hot spots we lack temporal series of biodiversity data (Dornelas et al., 2018;Newbold et al., 2020), hindering our ability to quantify the patterns of variation in species richness and composition needed to identify generalities in the responses of biodiversity to global change.
Drivers of environmental change, such as climate and land-use change, can influence temporal variation in local communities (i.e., species interacting in a given place during a given time period), where changes are generally quantified in terms of species richness and community composition (Magurran & Henderson, 2010;Newbold et al., 2020). A change in species richness simply results from the addition or reduction in the number of species. A change in community composition can be a consequence of taxonomic turnover, where the identity of co-occurring species change, and/or taxonomic reordering, where the identity of the species persists but their relative abundances change (Avolio et al., 2015;Collins et al., 2020;Dornelas et al., 2013) (Figure 1). Taxonomic turnover results from the immigration of new species into a community, or from local extinctions (Buckley & Jetz, 2008;Shurin, 2007). Taxonomic reordering occurs when the evenness in the distribution of species abundance in a community changes (Heard et al., 2012;Song et al., 2019), or when the pattern of rank abundances of species changes (Collins et al., 2020;Jones et al., 2017) (Figure 1). Understanding the temporal change in local communities requires unraveling the influence of richness, turnover, and reordering in the dynamics of communities (Avolio et al., 2015).
The few long-term studies of changes in biological communities carried out in the tropical region generally agree with the world-wide pattern of biodiversity change (Blowes et al., 2019; but see Beaudrot et al., 2016). For instance, a reordering of tree species over 35 years was detected in a network of study plots in the Amazonia forest lowlands (Esquivel-Muelbert et al., 2019). Dung beetle communities in the lowlands of Costa Rica, studied over a period of 30 years, presented a decline in species richness and taxonomic reordering (Escobar et al., 2008). A 20-year decline in the species richness of mammals was detected in tropical Queensland, Australia (Laurance et al., 2008). For birds, a 13-year study in the Amazonia lowlands found taxonomic reordering (Blake & Loiselle, 2016). However, longterm studies in tropical ecosystems are still scarce, in particular in the highly threatened, but poorly studied tropical Andes (Blowes et al., 2019).
Protected areas in the tropical region play a pivotal role in the conservation of biodiversity (Bruner et al., 2001;Naughton-Treves et al., 2005), but their effectiveness has been challenged (Laurance et al., 2012;Malhi et al., 2014). The biodiversity that such areas seek to conserve might be governed by metapopulation and metacommunity processes operating at scales larger than the protected area itself (Hansen & DeFries, 2007;Laurance et al., 2012). Reserves are embedded within a larger ecosystem, and they are not isolated from the effects of global climate or regional land-use change (Hansen & DeFries, 2007;Maiorano et al., 2008); as a result, there is a global concern about changes in taxonomic richness and community composition within protected areas (Ponce-Reyes et al., 2012;Velásquez-Tibatá et al., 2013).
Cajas National Park and the nearby Mazán reserve, located in southern Ecuador, harbor a representative avifauna of highelevation tropical Andean ecosystems and have been recognized as an important bird conservation area (Astudillo et al., 2015;Freile & Santander, 2005). Previous work in this area reported a decline in bird species richness in native and introduced forest habitats F I G U R E 1 Schematic diagram of potential community responses to the effects of environmental change of the Mazán reserve between 1994 and 2007 (Latta et al., 2011).
However, these results were based on a comparison between two points in time (1994-1995 and 2006-2007), covering a sampling interval of 12 years; nonetheless, communities can have heterogeneous changes over time, and data derived from continuous time series are better suited for revealing the processes underlying community dynamics (Reinke et al., 2019;Stouffer et al., 2021;White, 2019).
In this study, we explored avian diversity change over a period of 11 years in Cajas National Park and Mazán reserve, using data that have been collected continuously from 2006 to 2016 in three habitats: native forest, introduced forest (i.e., forests planted with exotic trees), and native shrub. Specifically, we analyzed changes in local communities to ask the following three questions: (1) Is taxonomic richness changing over time? Given the previous evidence of decline in richness in the study area (Latta et al., 2011), we expect a decline in richness across all habitats; (2) Is species composition changing over time? We expect a temporal decline in composition similarity across different habitats, corresponding to the global trend of local community composition change over time (Blowes et al., 2019;Dornelas et al., 2014); (3) Is the change in species composition a result of turnover and/or the reordering of species (i.e., changes in abundances)? Given earlier results from Latta et al. (2011), we expect change will be driven by both reordering and turnover. Our moderately long-term study offers an opportunity to reveal how communities are responding to the global environmental change in a biodiversity hot spot.

| Study area and study design
This study was conducted in Cajas National Park and the adjacent Mazán private reserve in the high Andes of Azuay province, Ecuador ( Figure 2). These areas are of global importance for biodiversity conservation, have been designated an Important Bird Area (IBA) (Freile & Santander, 2005), and are part of the "Macizo del Cajas" International Biosphere Reserve (UNESCO). The study area receives 1200-1500 mm of rain with a main rainy season from January to June, a dry season from July to September, and a secondary rainy season from October to December (Celleri et al., 2007). Mean monthly temperatures range from 5 to 12°C, but temperature variation within a day can range from 0 to 20°C (Celleri et al., 2007).

| Sampling protocol
At each sampling station, we captured birds by placing 20 mist nets (12 m × 32 mm mesh) along edges of narrow trails (1 m approximately) in fixed locations for two consecutive days ( Figure 2). Nets were open from dawn to 17:00 of day 1 and dawn to 11:00 of day 2. All birds were uniquely banded with a numbered aluminum metal band to identify recaptured individuals. Birds were sampled three times annually from 2006 to 2016 at each sampling station. To cover the climatic seasonality of the study area, we sampled once in the main wet season (within April 1-May 15), once in the dry season (within July 15-August 15), and once in the secondary rainy season (within November 9-December 18).
It is suspected that birds may sometimes learn the locations of nets over time and avoid being captured (Karr, 1981); however, avoidance of nets by previously captured birds likely lasts for only a few days (Ballard et al., 2004;Nur & Geupel, 1993). Therefore, we limited net avoidance biases by sampling only during two continuous days, with a minimum of 60 days between samplings periods.
The total number of sampled hours varied somewhat among sampling periods because we did not operate nets during rainy conditions; therefore, to consider variation in the sampling effort during each sampling period, both species richness and abundance were expressed as number of species or individuals captured in 100 mist net hours, where one mist net opened for 1 hr = 1 mist net hour.

| Data analysis
We obtained annual taxonomic richness and abundance based on the three field samples collected each year; we used this as our unit for all analyses. Annual species richness was obtained by calculating the cumulative species richness in 100 mist net hours per year, and annual abundance for each species was estimated as its average abundance in 100 mist net hours among the three sampling periods per year. We used band numbers to avoid double-counting birds within the sampling periods.

| Across-time comparisons
To determine whether species richness changed over time, we used linear models, with a normal distribution of residuals, for each habitat, with observed annual taxonomic richness as the response variable and time as the predictor variable. Species richness can be sensitive to variation in the detectability of species among samples (Boulinier et al., 1998); thus, we calculated detectability in species richness for each year in each habitat following Kéry and Schmid (2005), with the formula D = C/N, where C is the annual richness and N represents the potential richness. Potential richness (N) was estimated with a Jackknife estimator, obtained by using observed species richness of each of the three sampling periods within each year (Kéry & Schmid, 2005;Walther & Moore, 2005). Jackknife is a nonparametric richness estimator, based on a resampling technique that accounts for variation in detectability by leaving out one sample at a time and estimating averages over the rest of samples (Heltshe & Forrester, 1983). We explored if there was temporal variation in detectability in species richness over time with a linear regression.
We calculated the temporal rate of change in community composition in each habitat using time-lag analysis (Collins et al., 2000).
In time-lag analysis, a measure of community dissimilarity among sampling periods at increasing time lags is calculated. For the eleven years of study, we obtained ten one-year time lags (year 1 vs year 2, year 2 vs year 3, … year 10 vs year 11), nine two-year time lags (year 1 vs year 3, year 2 vs year 4, … year 9 vs year 11), through one tenyear time lag (year 1 vs year 11). Then, a linear regression between dissimilarity and time lags was used to evaluate whether there is directional change in community composition based on the slope of the regression (Collins et al., 2000). A matrix of species abundances recorded during each year was constructed for each habitat, from which we calculated dissimilarity in community composition among years using Euclidean distance. Euclidean distance was used because it incorporates information on joint absences between samples, and has clear geometric properties (Anderson et al., 2011;Collins et al., 2000).
Rare species can have a strong influence on dissimilarity measures, and those species could have been under-sampled in our field method. We, therefore, repeated the time-lag analysis including only dominant species, defined as the minimum number of species that together represent 50% of all individuals that are part of the assemblage of interest, which in our case was the annual assemblage each year for each habitat. That percentage (50%) has been previously shown to follow abrupt changes in the abundance curve of species found in Amazonian communities (Steege et al., 2013).
We further assessed if changes in species composition were the result of taxonomic turnover or taxonomic reordering. To explore the role of taxonomic turnover, we performed a time-lag analysis including only species presence or absence in the dissimilarity distance calculations. If taxonomic turnover among temporal samples is important in a community, a significant rate of change over time should be detected (Jones et al., 2017). Two aspects of taxonomic reordering were evaluated: evenness and rank abundance shifts. To test for changes in evenness over time, we created linear regression models using the annual Pielou's index of evenness as a response variable and time as the predictor variable. Pielou's evenness captures F I G U R E 2 Map of the study area (Google, 2019) showing the location of Llaviuco valley (A) in Cajas National park and Mazán reserve (B). Black dots shown in the inlets A and B represent the location of the mist nets in each sampling station the amount of evenness of a community relative to the maximum evenness possible for a given richness (equal abundance for all the species in the community) (Tuomisto, 2012). Rank abundance shifts were evaluated by measuring mean rank change in species across time (Avolio et al., 2019). Mean rank change calculates the degree of species reordering over time by comparing the amount of change in species ranks between continuous time periods (Avolio et al., 2019).
Lastly, to visualize changes in the communities over time we generated annual rank abundance distribution plots (Matthews & Whittaker, 2015), which depict different structural properties of assemblages such as changes in richness, evenness, and rank abundances (Matthews & Whittaker, 2015). To recognize the identity of the dominant species in each assemblage and to explore changes in species rank abundances over time, we used heat maps depicting the proportional abundance of dominant species.
The performance of all of the linear models described above was validated by visually exploring the distribution of residuals and homogeneity of variances; in all cases, the distribution of residuals indicated normality, and plots of residuals against fitted values reveal homogeneity of variances. Autocorrelation function plots (ACF) were used to check for potential temporal autocorrelations, and they did not indicate temporal autocorrelation of residuals. All of the analyses were performed using software R (R Development Core Team, 2020).

| RE SULTS
Overall, we captured 68 species across all habitats, with 50 species occurring in native forest, 57 species in native shrubs, and 49 species in the introduced forest (   Figure S1).
There was no temporal trend in the rate of taxonomic turnover in any of the habitats (Figure 4a, Table 2). Moreover, species evenness declined in the native forest, but not in the other habitats (Figure 4b, Table 2). Mean rank change was variable across habitats, without any particular temporal trend (Figure 4c).

Rank abundance distribution plots over time in each habitat
showed that in native forest and native shrub habitats, the propor-

| DISCUSS ION
Using a unique dataset from the tropical Andes, we examined how species richness and community composition of birds changed over 11 years in three habitats located in high-elevation protected areas.
Taxonomic richness remained constant in native forest, shrub, and introduced forest habitats. However, we detected taxonomic reordering in the native forest, where evenness of the distribution of species decreased over the studied period. This trend, resulting in increased dominance of some species, while others became rare in the native forest, is of conservation concern because it indicates that biodiversity could be changing over time even in the absence of local disturbances. Changes in avian diversity in the absence of local disturbances have been found previously in other tropical ecosystems (Blake & Loiselle, 2016;Escobar et al., 2008;Sigel et al., 2006;Stouffer et al., 2021). The continuation of our monitoring efforts will reveal if these changes in evenness will result in local extinctions or other detrimental losses. Our findings in the native forest correspond to those of several other studies where species richness remained constant over time while changes were recorded in the identity of species in the avian community, and/or their abundances (Blowes et al., 2019;Thibault et al., 2004).

Constant species richness over time across all habitats agrees
with global analyses that suggest that at the local scale species richness is relatively stable (Blowes et al., 2019;Dornelas et al., 2014;Thibault et al., 2004). However, these results contrast with those of a previous study in the Mazán Valley which was based on two temporal samples, 1994-1995 and our first sampling period in 2006-2007. In that analysis, richness appeared to decline in native and introduced forests (Latta et al., 2011). This difference may be the result of stochastic variation of communities that were sampled at two different points in time (Fournier et al., 2019), or differences in sampling protocols as compared to the present study. Alternatively, given extensive deforestation in the region (White & Maldonado, 1991), regionally rare species may have already been absent when we started sampling, and this may have affected estimated species richness. This possibility is supported by Latta et al. (2011) who found that those species present in 1994-1995 and absent 11 years later were considered to be rare in Cajas National Park. Hemispingus TA B L E 1 List of bird species captured with mist nets in native forest, native shrubs, and exotic forest habitats in Cajas National Park and Mazán reserve, southern highlands of Ecuador  tions (Jackson & Sax, 2010;Stegen et al., 2013). This contrasts with populations in disturbed habitats that are often transient, leading to high levels of temporal community composition change (White et al., 2010). For instance, across habitats comparisons have shown that temporal composition change is higher in disturbed or successional environments compared to undisturbed habitats (Kampichler et al., 2014;Stegen et al., 2013). suggesting that local vegetation dynamics have not strongly influenced observed changes in assemblage evenness. As local assemblages are highly connected to dynamics of the regional species pool, it is possible that changes in the native forest are a result of land-use change operating beyond the boundaries of the Mazán reserve (Latta et al., 2011). This is difficult to evaluate as there are no updated fine-scale estimates of land-use change in the region.
Transformation of mountain forest to pastures is an ongoing process in the Ecuadorian Andes (Ministerio del Ambiente, 2016), and habitat conversions to promote grazing continue to alter the composition of bird assemblages (Palacio et al., 2020;Santillán et al., 2019;Tinoco et al., 2018). Finally, tropical mountains are highly sensitive to climate change (Urrutia & Vuille, 2009), and recent evidence suggests range shifts and even extirpations of birds in the tropical Andes as a result of climate change impacts (Feeley et al., 2011;Freeman et al., 2018). Detailed regional information on land conversion rates and climate change, combined with broader monitoring of species composition across space, is required to better explore the driving factors of biodiversity change at local scales as seen here.
Temporal changes in evenness signal an assemblage that is undergoing a transitional phase (Smith et al., 2009), usually triggered by a disturbance in the environment (Dornelas, 2010;Kim et al., 2013;Matthews & Whittaker, 2015;Shurin, 2007). If evenness continues to decline in the native forest, it is possible that local extinctions might occur and the assemblage may reach a new species equilibrium with only a subset of the original species (Jackson & Sax, 2010;Smith et al., 2009;Supp & Ernest, 2014 (Laurance et al., 2012). Because of a lack of habitat-level replicates, our study cannot be used to F I G U R E 5 (a) Annual rank abundance distribution plots of the bird assemblage in native forest, shrub, and introduced forest from 2006 to 2016 in Cajas National Park and Mazán Reserve in the highlands of southern Ecuador. The y-axis represents the proportion of abundances of each species, while the x-axis represents the rank of each species. Each color line represents a different year, and within years, species are ordered from the species with the highest rank to the species with the lowest rank. (b) Heat maps depict the proportions of the dominant species within assemblages from 2006 to 2016 in each habitat. The y-axis represents the proportion of abundances of each species, while the x-axis represents species names. Codes of species names correspond to complete names presented in Table 1 quantitatively compare rates of community change among habitats; we acknowledge that limitation. Yet, it is striking that the less altered habitat, native forest, has a higher rate of composition change than the more altered shrub and introduced forest habitats. Moreover, the native forest in Mazán reserve is managed under strict protection practices; thus, our results suggest that well-protected reserves are not immune from the current global change. Temporal declines in richness, and changes in community composition, have also been observed in other strictly protected areas of the tropical region, bringing into question whether protected areas are sufficient for maintaining biological diversity in the long term (Laurance et al., 2012;Malhi et al., 2014;Stouffer et al., 2021). Unfortunately, there are very few studies documenting temporal trends in Andean biodiversity, even in well-studied groups such as birds. Our study highlights the value of long-term studies in the tropical Andes, as we show that species composition of birds in a montane forest is changing, providing additional support for global trends in biodiversity change (Blowes et al., 2019;Dornelas et al., 2014;Newbold et al., 2020).

ACK N OWLED G EM ENTS
We want to thank Francisco Sanchez, Juan Carlos Quezada, José Cáceres, and people from the "Department of Gestion Ambiental" ETAPA for logistical support in Cajas National Park. We want to thank specially the multiple students from University of Azuay who have been part of data collection along these years, in par-

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available in the Dryad Digital Repository: https://doi.org/10.5061/dryad. r2280 gbds (Tinoco et al., 2021).