Local contributions to beta diversity in urban pond networks: Implications for biodiversity conservation and management

An understanding of how biotic communities are spatially organized is necessary to identify and prioritize habitats within landscape‐scale biodiversity conservation. Local contribution to beta diversity (LCBD) identifies individual habitats that make a significant contribution to beta diversity and may have important practical implications, particularly for conservation of habitat networks. In this study, we develop and apply a conservation prioritization approach based on LCBD in aquatic invertebrate communities from 132 ponds.


| INTRODUC TI ON
One of the primary goals of ecological research is to understand spatial patterns of taxonomic richness and community composition to develop effective and ecologically relevant conservation strategies and protected area networks (Socolar et al., 2016). Ecological diversity can be divided into alpha, beta and gamma components (Whittaker, 1960), each of which offer complimentary perspectives on the spatial organization of biological communities. Interest in beta diversity (defined here as the variation in biotic communities within a defined geographical region: Whittaker, 1960) and its application in ecological conservation has increased significantly in recent years (Socolar et al., 2016). Beta diversity can be assessed by calculating taxonomic dissimilarities among ecological communities and can be separated into two distinct components: species replacement (turnover) and richness difference (nestedness; Baselga et al., 2012;Legendre, 2014). Species replacement refers to the turnover of species among sites and may reflect species gained or lost due to historical events, competition and/or environmental sorting (Baselga, 2010;Legendre, 2014). Richness difference reflects how communities differ from each other in their taxonomic richness.
Taxa present at sites with low richness may comprise a subset of taxa from more species-rich sites which may reflect habitat heterogeneity or ecological processes that promote species thinning (Legendre, 2014).
Recent developments in beta diversity analyses allow an assessment of the extent to which individual sites contribute to total beta diversity within a given region (Legendre & Cáceres, 2013). The local contribution to beta diversity (LCBD) metric quantifies the ecological uniqueness of each site within the context of a wider ecosystem network, with high values for a given site indicative of high dissimilarity between the community and other sites in the region (Heino & Grönroos, 2017). Further, LCBD can be partitioned into the individual site contribution of the species replacement ( Repl LCBD) and richness difference RichDiff LCBD components (Legendre, 2014). By identifying the individual sites that contribute most to beta diversity, significant ecological advances can be made in understanding spatial patterns and the processes that drive them. A limited number of studies have examined LCBD and ecological uniqueness among freshwater communities in recent years, particularly in relation to their environmental correlates Heino & Grönroos, 2017;Tonkin et al., 2015). However, the environmental variables that influence the contribution of replacement and richness difference to total beta diversity remain poorly understood (but see Hill, Biggs, et al., 2017;, and knowledge regarding local contributions to beta diversity among pond habitats is absent. In addition to advancing ecological theory, LCBD approaches (and the replacement and richness difference components) have important practical implications, particularly for conservation planning of habitat networks (Vilmi et al., 2017). Understanding how communities are spatially organized (and what drives those patterns) can facilitate more effective biodiversity conservation strategies at a landscape (gamma) scale. For example, patterns of richness difference among communities suggest conserving individual species-rich sites should be a priority given that other sites effectively represent subsets of the most species-rich sites.
However, high species replacement would suggest conserving a range of sites with contrasting biota would be more appropriate (Hill, Biggs, et al., 2017;Hill, Heino, et al., 2017;Socolar et al., 2016). Currently, pond conservation is heavily focussed on conserving individual sites of high taxonomic richness, despite recent evidence that landscape-scale pond conservation can be more effective in supporting the greatest biodiversity (Hassall et al., 2011;Hill, Biggs, et al., 2017;Hill, Heino, et al., 2017). To be considered for conservation in many regions, pond habitats need to meet specific criteria, for example, in the UK ponds are required to support several rare species, recording >50 aquatic invertebrate species and/or a Predictive System for Multimetrics score of >75% (BRIG, 2008). However, it could be that conserving ecologically unique (sites of greatest beta diversity) sites may significantly increase the proportion of the regional species pool that is protected, thereby facilitating landscape-scale conservation. In addition, LCBD approaches may identify sites of conservation importance that fall outside of the current criteria utilized for pond conservation. High LCBD values may reflect unusual species combinations and/or environmental conditions, the presence of rare taxa and therefore sites of potentially high conservation value (Heino & Grönroos, 2017). High LCBD values may also indicate sites of low taxonomic richness (high richness difference) that are suitable for restoration (Legendre, 2014). Identifying effective approaches for biodiversity conservation are required as anthropogenic activities continue to threaten global biodiversity and freshwater species loss continues (Sánchez-Bayo & Wyckhuys, 2019).
Pond networks provide ideal systems to study spatial ecological processes and nature conservation strategies as they are highly abundant and clearly delineated (local populations and communities are easily quantified) in the landscape, have a wide range of ecological and spatial (connectivity) gradients that are easy to quantify and are easy to sample in a replicable, quantitative and representative way (De Meester et al., 2005). The LCBD approach allows comparison among ponds identified as significantly contributing to LCBD with those ponds that would be considered to be of high taxonomic richness. In so doing, it is possible to explore the benefits (increased gamma diversity) of considering LCBD and richness individually, as well as in combination for the identification of pond sites for conservation. In this study, we use a large freshwater pond community dataset to explore pond conservation potential based on LCBD. We partition LCBD into richness difference and species replacement and identify, for the first time, key environmental variables that are associated with LCBD and local contributions of species replacement and richness difference among ponds.
Our analysis operates at two scales that are relevant to ecological and conservation planning: within urban settlements and nationally across England.

| Data management
Macroinvertebrate and environmental data from ponds in five urban settlements across England (Halton-a local government district in Cheshire: n = 25, Loughborough: n = 41, Stockport: n = 16, Birmingham: n = 30 and Huddersfield: n = 20, total = 132) were compiled from previous studies (Gledhill et al., 2008;Hill et al., 2015;Pond Life Project, 2000;Thornhill, 2013;Wood et al., 2001). Ponds are defined here as lentic water bodies <2ha in area that hold water for at least 4 months of the year (Williams et al., 2010). For the location of urban pond sample sites, see Figure S1. Full details and the rationale for data management procedures are presented in ,  and summarized here.
In this study, urban settlements were defined as areas containing >10,000 people and being >20 ha in area (UKNEA, 2011). Developed land use areas (DLUA's) were used in this study to define a pond as urban (see supplementary material by Hill, Biggs, et al. (2017) for a discussion on urban pond definitions). The urban ponds studied here were located along an urban settlement gradient, from a densely populated city (Birmingham: >1 million inhabitants) to a smaller, less densely populated town (Loughborough: ~60,000 inhabitants).
The ponds were located in domestic gardens, industrial sites, urban green space, adjacent to roads and in commercial districts.
Aquatic macroinvertebrate sampling methods undertaken by the five urban pond studies followed two methodologies: (1) a single 3min sweep sample in each urban pond, divided between the microhabitats present (Loughborough, Huddersfield and Birmingham) and (2) exhaustive sampling of macroinvertebrates in all available microhabitats until no new species were recorded (Halton and Stockport).
However, taxonomic richness recorded using the two sampling strategies has been demonstrated to be comparable (see . Pond area, pH, altitude, pond shading (estimated percentage of the pond surface shaded by trees and scrub) and pond emergent macrophyte coverage (EM: estimated percentage of pond covered with emergent macrophytes) were recorded in most studies and examined here. However, pond shading was not recorded from ponds in Huddersfield and pH was not recorded from ponds in Stockport (see ,   Table 2, for a summary of the environmental characteristics of ponds in the urban settlements). The percentage of urban land coverage within a 250 m buffer of each pond was also calculated, to account for the level of urbanity around each pond (Table S1). Land cover data around each pond were derived from the Centre for Ecology and Hydrology National River Flow Archive (https://nrfa.ceh.ac.uk/conte nt/land-use). The invertebrate datasets collated from the contributing studies were converted into a presence-absence matrix to remove any sampling bias. Most macroinvertebrates were identified to species level, Diptera were resolved to family level, Bithynia, Argulus and Hydroptila to genus level, and Oligochaeta, Sphaeriidae, Collembola and Hydrachnidiae were recorded as such. It is acknowledged that the richness recorded in this study has been underestimated given that the highly diverse Diptera group were only resolved to family level.

| Data analysis
All statistical analyses were undertaken in R (R Core Team, 2019). We first calculated the pairwise total beta diversity and the contribution of species replacement and richness difference components to total beta diversity (based on Baselga family, Jaccard-based indices) using the function beta.div.comp from the adespatial package (Dray et al., 2019) for the entire dataset and individual urban settlements.
We then calculated the local contribution to beta diversity (LCBD; using Hellinger transformed presence-absence data) for each site across the entire dataset and among sites in each urban settlement separately using the function beta.div in the adespatial package (Legendre & De Cáceres, 2013). The statistical significance of individual site contributions (individual LCBD values) can be calculated through random, independent permutations within the community matrix (Legendre & Cáceres, 2013). The local contributions of species replacement ( Repl LCBD) and richness difference ( RichDiff LCBD) to total beta diversity (based on Baselga family, Jaccard-based indices) were calculated using the function beta.div.comp (Legendre, 2014).

Differences in taxonomic richness, LCBD, Repl LCBD and
RichDiff LCBD values among ponds from the five urban settlements were assessed using a Kruskal-Wallis test (using the kruskal.test function). Nemenyi post hoc tests, using the function posthoc.kruskal.nemenyi.test in the PMCMR package (Pohlert, 2018), were undertaken to determine where significant differences among the five urban settlements occurred. The response of LCBD, Repl LCBD and RichDiff LCBD values (dependent variables) was explored in relation to the taxonomic richness (independent variable) via separate sets of statistical models across the entire dataset and each of the five individual urban settlement datasets (n = 18; 3 responses × 6 datasets).
Each statistical set modelled the influence of taxonomic richness via 4 statistical functions-linear, quadratic, exponential and logarithmic (sensu Fornaroli et al., 2019). For statistical sets performed on individual settlement datasets, traditional regression models (y = αx + β + ε) were employed. A linear mixed-effect model design (y = αx + β + (1 | City) + ε) was used on the entire dataset, whereby taxonomic richness was used as a fixed effect (as above), and each settlement was used as a random effect (to account for potential spatial autocorrelation) using the lmer function in the lme4 package (Bates et al., 2019). For each statistical set (each containing the same dependent variable vs. taxonomic richness modelled via the 4 statistical functions), statistical summaries of the model exhibiting the lowest AIC value were reported. For traditional regression models (i.e. individual settlements), the partial R 2 for each predictor variable was calculated for the final models using the function etasq from the heplots package (Fox et al., 2018). For linear mixed-effect models (entire dataset), partial R 2 values were derived using the r2beta function in the r2glmm package (Jaeger, 2017). For all traditional regression and linear mixed-effect models, residuals were plotted against fixed values to assess the homogeneity of variances and identify outliers (up to 3 data points were removed), while quantilequantile plots were inspected to ensure that models were normally distributed. For individual urban settlements, we analysed the degree of spatial autocorrelation in the response variables based on Moran's I values using the correlog function in the pgirmess package (Giraudoux et al., 2018) to ensure the assumptions of independence were valid (see Supplementary Material 2). Differences in the total richness and number of unique species between significant LCBD sites and sites which recorded >50 taxa were also examined. Sites supporting >50 taxa were selected as this reflects the taxonomic diversity required for a pond to be considered for conservation in the UK. Total (gamma) richness was calculated for (1) the significant LCBD sites, (2) sites containing >50 taxa and (3) significant LCBD sites and sites containing >50 taxa combined across the entire dataset, as well as each urban settlement individually. To identify whether these three methods resulted in greater gamma diversity than would occur by chance, for each individual urban settlement and the entire dataset, we randomly selected the same number of sites as identified within each of the three methods mentioned above. These sites were randomly sampled without replacement, and this process was repeated across 1,000 simulations to produce each comparison. For each randomly selected dataset, the mean total (gamma) taxonomic richness (and standard deviation) was calculated. To examine whether the average total richness recorded from a random selection of ponds was less than the total richness recorded from significant LCBD sites and sites with >50 taxa, a one sample t test was performed.

| LCBD, Repl LCBD and RichDiff LCBD values among ponds from the 5 urban settlements
In total, 338 macroinvertebrate taxa were recorded from ponds across the five urban settlements: Halton (108 taxa, median: 28), RichDiff LCBD values were significantly higher (p < 0.01) in Birmingham and Loughborough than Halton, Stockport and Huddersfield ( Figure 2c). When the urban settlements were considered individually, 6 ponds in Halton, 5 in Loughborough, 2 in Stockport, 6 in Birmingham and 2 in Huddersfield significantly contributed to total beta diversity (Table 1).

| Contribution of significant LCBD sites to regional taxonomic richness
A total of 148 taxa and 21 unique taxa were recorded from the 21 significant LCBD ponds across the entire dataset (Table 1; Supplementary Material 3). Ponds that supported >50 taxa recorded a total of 240 taxa and 43 unique taxa when the entire dataset was considered. When significant LCBD ponds and those with >50 taxa were considered together, a total of 279 taxa and 71 unique taxa occurred across the entire dataset (Table 1). The total number of species and unique taxa recorded from the 5 urban settlements for significant LCBD ponds, ponds that support >50 taxa and both combined are reported in Table 1. A total of 31% of ponds were recorded as significant LCBD sites and/or sites with >50 taxa when all sites were considered, 66% in Birmingham, 20% in Loughborough, 10% in Huddersfield, 37% in Stockport and 24% in Halton.
The three methods to assess gamma diversity conservation potential, that is (1) significant LCBD sites, (2) sites containing >50 taxa and (3) significant LCBD sites and sites containing >50 taxa combined were applied to each of the six datasets and compared against an equal number of randomly selected samples (1,000 simulations).
The total richness recorded among a random selection of sites was greater than the total richness recorded from significant LCBD sites when the entire dataset was considered (mean random richness: 197.6 ± 0.48 vs. total significant LCBD richness: 141; t = 117.68, Halton (71 ± 0.23 vs. 41; t = 132.58, df = 999, p > 0.05). However, the total richness of randomly selected sites was significantly less than the total richness recorded from significant LCBD sites in Huddersfield (26.26 ± 0.21 vs. 30; t = −17.69, p < 0.05).
Total taxonomic richness was significantly less among the randomly selected ponds compared to sites with high taxonomic diversity (sites with >50 taxa) when the entire dataset was con- The total richness of randomly sampled sites was significantly less than the total richness recorded when significant LCBD sites and sites with high taxonomic diversity were combined and considered together (mean random richness: 249.14 ± 0.36 vs. significant LCBD and high taxonomic diversity sites total richness: 271; t = −60.97,
When the individual urban settlements were considered, the final regression model explained 38% of the variation in LCBD in Halton (adj. R 2 = 0.38, F = 3.07, p < 0.05; significant predictor-EM) and 49% in Stockport although the final model was not significant (adj. R 2 = 0.49, F = 3.02, p = 0.07: EM) ( Table 2). For pond LCBD in Loughborough, the final model had poor explanatory power, explaining 3% of the statistical variation (adj. R 2 = 0.03, F = 1.22, p = 0.32: EM) ( Table 2). Among ponds in Birmingham, 11% of the variation in LCBD was explained by the final regression model (adj. R 2 = 0.11, F = 1.54, p = 0.21: surface area). A total of 15% of the variation in Repl LCBD were explained by the final regression model in Halton (adj. R 2 = 0.15, F = 1.63, p < 0.05: EM) and 50% in Stockport, although the final model was not significant (adj. R 2 = 0.50, F = 3.15, p = 0.07; EM). Surface area was a significant predictor of variation in RichDiff LCBD in Halton, while emergent macrophytes were the only significant predictor for RichDiff LCBD in Stockport. However, the TA B L E 1 Macroinvertebrate richness and unique taxa (taxa only recorded from those ponds) recorded among significant LCBD sites, sites with >50 taxa and both combined. Total number of ponds significantly contributing to LCBD or recording >50 taxa are presented in parenthesis No environmental variables were significant predictors of pond, LCBD, Repl LCBD or RichDiff LCBD in Huddersfield (Table 2). In addition, no significant predictors were associated with Repl LCBD and RichDiff LCBD among ponds in Loughborough or Birmingham.

| D ISCUSS I ON
This study is the first to examine the contribution of individual urban pond sites to patterns of overall beta diversity (LCBD), richness difference ( RichDiff LCBD), species replacement ( Repl LCBD), the environmental variables influencing those patterns and their relevance and potential application for biodiversity conservation at a national and city scale. Our results indicated regional variability in the association of taxonomic richness and ecological uniqueness (LCBD) between the urban settlements examined in this study.
Ecological uniqueness was negatively related to macroinvertebrate taxonomic richness when all ponds were considered, and among ponds in Birmingham, Stockport and Halton, but no association was identified among ponds in Huddersfield and Loughborough. Sites with high taxonomic diversity may exhibit low LCBD values due to the increased chance of sharing species with other sites in the region (Szabó et al., 2019). A negative association between LCBD and taxonomic richness has been previously recorded among fish (Legendre & Cáceres, 2013), macroinvertebrates (Heino & Grönroos, 2017),

F I G U R E 3
Relationship between the local contribution to beta diversity (LCBD) and taxonomic richness among ponds across the entire dataset (a) and within individual urban settlements:  (Szabó et al., 2019;Vilmi et al., 2017), microbial communities (Teittinen et al., 2017), dung beetles (da Silva et al., 2018) and tree species (Qiao et al., 2015). However, some studies have recorded a positive association between LCBD values and taxonomic richness (e.g. fish assemblages- Kong et al., 2017) or no association (e.g. diatoms in lake ecosystems- Vilmi et al., 2017). A positive relationship may reflect the colonization of novel species (e.g. migratory species, rare species) in communities (da Silva et al., 2018). Qiao et al. (2015) identified a negative association when common tree species were considered but a positive association when only rare species were considered. Therefore, the occurrence of rare and common species in freshwater communities may play a significant role in determining whether the LCBD-taxonomic richness association is positive, negative or non-significant (Cucherousset et al., 2008;da Silva et al., 2018). A wide range of processes may be influencing ecological communities, resulting in different associations between LCBD and taxonomic richness including; primary productivity, environmental filtering, hydro-climatological processes, species interactions, species dispersal, disturbance regimes, stochastic processes and biotic interactions (Bennett et al., 2009;Heino & Grönroos, 2017;Tonkin et al., 2015). Furthermore, the patterns found between LCBD and taxonomic richness may be influenced by the type of data used (Szabó et al., 2019). For example, a negative association between LCBD and taxonomic richness among stream macroinvertebrates F I G U R E 4 Relationship between the Repl LCBD and RichDiff LCBD among ponds across the entire dataset (a), and within individual urban settlements: Halton (b), Loughborough (c), Stockport (d), Birmingham (e) and Huddersfield (f). Significant LCBD sites are presented in bold

Stockport Loughborough
Halton was recorded when presence-absence data were used; however, a non-significant relationship was recorded when abundance-based data were used (Heino & Grönroos, 2017). Our results reflect and support the findings of previous research highlighting the complex relationship between taxonomic richness and LCBD across biological groups, geographical regions and different ecosystems.
Across the entire dataset and in four of the five urban settlements studied, pond sites that contributed most to total beta diversity (significant LCBD sites) were characterized by high replacement and the lowest richness difference values, indicating that there were high numbers of macroinvertebrate taxa recorded in these ponds that were not recorded in the other ponds (unique species).
This finding may be due to species replacement typically being positively correlated with total beta diversity, while richness difference commonly displays a negative correlation with total beta diversity (Soininen et al., 2018). This pattern also indicates that species replacement broadly measures the same phenomenon as total beta diversity (the compositional variation between sites; Soininen TA B L E 2 Significant predictors of taxonomic richness, local contribution to beta diversity (LCBD) and the replacement ( Repl LCBD) and richness difference ( RichDiff LCBD) components of LCBD across the entire dataset and for each urban settlement, resulting from regression analyses. The statistical function employed for each significant environmental variable is presented in parenthesis   (Legendre, 2014;Picazo et al., 2012).
Ponds from the five urban settlements studied showed considerable variability in the effects of environmental variables on LCBD.
Similarly, Tonkin et al. (2015) found significant variation in the relationship between LCBD and environmental variables across five different invertebrate stream datasets. When the entire dataset was considered and at a city scale, pond surface area and emergent macrophyte cover were the most important predictors of LCBD, Repl LCBD and RichDiff LCBD. This suggests they are key variables governing ecological uniqueness and have been reported in previous studies to be key predictors of compositional variation in ponds (Hassall et al., 2011;Hill et al., 2019). Interestingly, the percentage urban land coverage around a 250m buffer of each pond was not found to be an important driver of LCBD, Repl LCBD and RichDiff LCBD when the entire dataset was considered or when the individual urban settlements were considered, contrasting with the findings of . However, increasing urban land cover was found to be a significant negative predictor of taxonomic richness when the entire dataset was considered and among ponds in Loughborough, a finding which has been documented in other aquatic and terrestrial faunal studies (e.g. Mckinney, 2008;Roy et al., 2003). Findings from this study suggest that the influence of local environmental variables on LCBD may override the effects of urban land cover; however, this explanation is speculative as there have been so few studies examining LCBD among urban freshwater habitats. LCBD, Repl LCBD and RichDiff LCBD were poorly explained by the local environmental variables in ponds from Huddersfield and Birmingham, supporting the findings in streams by Heino and Grönroos (2017).
The weak relationship between local environmental variables and LCBD in this study may reflect an impacted regional species pool from anthropogenic disturbance (Tonkin et al., 2015). However, the poor prediction of LCBD may reflect other unmeasured local environmental (e.g. fish presence/density, depth, conductivity, dissolved oxygen, nutrient concentration, submerged macrophyte cover and macrophyte complexity) and spatial variables (connectivity / isolation) that may be important drivers of ecological uniqueness. Future research is required to examine the possible effects of these unmeasured variables on LCBD among ponds. This context-dependent pattern between the urban pond datasets is a key challenge in developing theoretical understanding of the environmental predictors governing the ecological uniqueness of freshwater ecosystems (Tonkin et al., 2015).

| Conservation implications
In Europe, pond conservation is currently focused on protecting individual sites with rare species or with records of very high biodiversity . However, this study and others have demonstrated that ponds contribute most to diversity at the landscape-scale . High beta diversity among ponds could almost entirely be attributed to species replacement indicating that pond conservation would be most effective at a landscape-scale, protecting a wide range of pond sites (Hill, Biggs, et al., 2017;Hill, Heino, et al., 2017). However, given the limited resources available for biodiversity conservation and restoration, protecting large numbers of ponds across a landscape is often impractical, particularly in urban areas. Therefore, assessing the landscape-scale conservation value of individual sites effectively is particularly important to ensure the maximum possible biodiversity is protected and sustained with the limited funding available.
Protecting sites of high ecological uniqueness (significant LCBD sites) may provide an opportunity to focus efforts on landscapescale conservation, while not deviating from the current practice of designating individual ponds for conservation. LCBD goes beyond typical measures of beta diversity (an overall measure across the landscape) and provides a value of the contribution to beta diversity for individual sites (Heino & Grönroos, 2017). Significant LCBD sites in this study typically supported unique species (taxa only recorded in high LCBD ponds due to their high species replacement) had unique compositions of species and contributed to regional diversity within most of the urban settlements examined. However, the low taxonomic richness recorded among the significant LCBD sites may reduce their perceived importance to practitioners for conservation. If considered individually, an LCBD-based approach may not be suitable to ensure the maximum possible biodiversity is protected. However, when prioritizing significant LCBD sites and sites with high taxonomic diversity together (conserving both high taxonomic diversity and ecologically unique site) more taxa are protected than high taxonomic diversity sites alone or those sites that would be protected purely based on chance (random selection of ponds in the network). In this study, 70%-97% of the regional species pool was protected when ecologically unique and high taxonomic diversity sites (>50 taxa) were considered together compared to 54%-94% when only sites with high taxonomic diversity were considered. The LCBD metrics presented here are straightforward to incorporate into spatial planning exercises or standard conservation planning software (e.g. Zonation; Moilanen et al., 2009, Marxan;Ball et al., 2009) as an additional layer for parameterization. Indeed, complementarity of protected areas is a common principle in systematic reserve design (Margules & Pressey, 2000) but is rarely applied at smaller scales or for specific habitats (Önal & Briers, 2002). Including ecologically unique sites within current conservation practices alongside sites of high taxonomic richness: (1) effectively enables landscape-scale conservation to be undertaken, as those sites that contribute most to beta diversity are being protected; (2) continues to ensure sites of highest biodiversity are protected; (3) ensures a significant proportion of the regional richness pool is protected; and (4)  This study also highlighted variability between urban settlements, as significant LCBD sites in Birmingham recorded higher richness difference values, while significant LCBD sites in other urban settlements were characterized by a high species replacement. Among those urban areas (Loughborough, Halton, Stockport and Huddersfield) where pond networks are structured by species replacement, protecting ecologically unique and taxonomically rich sites will maximize the number of taxa supported. However, where there is a high contribution of richness difference to beta diversity, such as Birmingham, traditional conservation efforts could focus on sites of high taxonomic richness, rather than LCBD sites as these formed subsets of sites with high taxonomic richness and made a limited contribution to the regional species pool (>2%).
Significant LCBD sites and high diversity sites together constituted between 10% and 66% of the pond network in the urban settlements. Using LCBD as a compliment to taxonomic richnessbased conservation provides significant opportunity to protect the most important ponds sites and ensure the persistence of aquatic biodiversity in urban landscapes, without needing to protect every pond in the network. Further, many of the ponds identified as significant LCBD or high diversity sites were constructed for anthropogenic purposes (e.g. stormwater collection). This pond conservation method supports a significant proportion of the regional species pool and can provide a cost-effective (minimal logistical and financial input to achieve), multifunctional approach to urban nature conservation, while also ensuring that the primary function of the ponds is maintained (Oertli & Parris, 2019). Given the variability between urban settlements, prioritizing ponds for landscape-scale conservation should be examined at an urban settlement-scale, and the use of species replacement ( Repl LCBD) and richness difference ( RichDiff LCBD) metrics outlined in this study can facilitate regionally effective conservation planning decisions (da Silva et al., 2018). Given the predicted increase in future urban land cover (Seto et al., 2012), identifying effective conservation approaches is essential to ensure biodiversity is sustained and where possible enhanced, and detrimental anthropogenic effects are minimized. MJH wrote the manuscript. All authors reviewed manuscript drafts and gave final approval for publication.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13239.