Identifying ‘climate keystone species’ as a tool for conserving ecological communities under climate change

Climate change affects ecological communities via impacts on species. The community's response to climate change can be represented as the temporal trend in a climate‐related functional property that is quantified using a relevant functional trait. Noteworthy, some species influence this response in the community more strongly than others.

. At the highest hierarchical level, however, the sources of variation in the contributions of individual species to the community-level response are not as well understood.Under climate change, a community in a particular location has alternative trajectories from being stable (with little-to-no species turnover) to becoming an entirely different community (with complete species turnover) over time.The climate-driven changes in species' abundances will therefore mediate both the community structure and functions, and the consequent trajectory of the entire community.Some species are more responsive to changing environmental conditions and may adapt, move or go locally extinct at a higher rate than others (Bateman et al., 2019;Comte et al., 2014;Hällfors et al., 2021;Wu et al., 2018).Predictably, such highly responsive species should influence more strongly the overall response of the community.In the context of climate change, the species-specific contributions to the community-level response can be quantified by comparing long-term trends in the average climate response trait of communities, such as phenology, with and without the focal species.In particular, through quantification of species-specific contributions to the community-level response, it is possible to identify which species act as 'climate keystone species' and influence the communities most in proportion to their local relative abundance.
Importantly, weighting the contribution of a species to the local community's climate change response with its relative abundance allows disentangling species-specific effects beyond dominance.Our approach expands the earlier definitions of keystone species that highlighted the top predators (Paine, 1966) or the species triggering observable cascading effects (Cottee-Jones & Whittaker, 2012).
Rather, our definition of the climate keystone species follows a more general keystone species concept relating species' abundances to their effects on the ecosystem overall (rather than to other species specifically; Power et al., 1996) and considers their keystoneness as a continuous species' property in the context of the community's ability to adjust and persist under climate change.
Research on community-level responses to climate change has dual relevance.First, for conserving biodiversity at a broader scale of communities and ecosystems under climate change, quantitative syntheses of communities' responses are needed.Second, climate change may change ecosystem functions, such as carbon sinks and storage (Davidson & Janssens, 2006;McGuire et al., 2009), that are essential for ecosystems' stability, which is why the communities' functional responses are important to study directly.To effectively measure such ecosystem functions, relevant ecological traits for species' climate change response are needed.Although establishing trait-climate and trait-ecosystem functioning links is beyond the scope of this paper, we strongly stress that such links need to be confirmed case-specifically before climate keystone species can be identified.The links between climate change responses of ecological communities and ecosystem functioning can be either direct or indirect.For example, in previous studies, climate change responses have been linked to other traits (birds; Sandel, 2019) and changes in other community affinities, such as N, pH and light (plants;Becker Scarpitta et al., 2017), which showcases simultaneous trait-climate change and trait-ecosystem functioning links.More concretely, climate keystone species can accentuate (i.e.contribute positively) or dampen (i.e.contribute negatively) the temporal trend in the climate change response trait of the community.Some species may speed up the community's overall adaptation to the new environmental conditions, while others may maintain the community's functional properties even under the changing environmental conditions.If climate keystone species with accentuating and dampening effects on the community-level response differ in their functional properties (e.g.specialization-or life-history-related traits), their impact on the community extends beyond the climate change response contribution to the overall functionality of the community.Assessing the species' keystoneness in climate change context can provide the basis to prioritize the conservation of ecological communities and associated ecosystem functions alongside with the traditional protection of rare and endangered species (Riva & Mammola, 2021;Violle et al., 2017).
Here, we present a quantitative tool for identifying 'climate keystone species' that strongly accentuate or dampen the communitylevel response to climate change, disproportionately to their relative abundance in the local community (Figure 1).This tool brings together components of earlier research on trait-climate and traitecosystem functioning links and keystone species concept in a novel way.The tool provides a systematic, transparent and testable way of quantifying species' contributions to the community-level changes.
Our goal is to allow identifying potential climate keystone species that could then be subjected to further research and conservation attention.We also give guidance for how to classify the climate keystone species and select the most efficient way to apply the climate keystone species identification tool to conservation.Finally, we demonstrate the identification of the climate keystone species with North American bird community data.

| IDENTIF YING CLIMATE KE YS TONE S PECIE S
We present a stepwise tool for identifying climate keystone species, which is applicable across taxa and ecological systems (Figure 1).
In general terms, a species is a climate keystone species if it has a stronger effect on the observed response of the entire community to climate change over time than expected based on its relative abundance.The community-level response to climate change can be quantified using a community property of interest that is derived from species' functional traits.The climate keystone species can be considered when preserving a community's ability to continue functioning under climate change in cases where the climate keystone species either strongly dampens a negative trend or strongly accentuates a positive trend in the relevant functional property of the community.For example, in the context of food webs, we may be interested in identifying which species contribute strongly to the changes in the food web's overall phenology and thereby the spatiotemporal matching of pairwise resource-consumer interactions.
Climate change can strongly affect species' phenology (Hällfors et al., 2021;Roslin et al., 2021) and lead to phenological mismatches between interacting species (Kharouba et al., 2018), for instance such that the spring phenology of prey invertebrates is more closely tracking warming temperatures than that of the predator birds (Burgess et al., 2018).Following the differential ability to track climatic F I G U R E 1 Schematic illustration of the five steps to identify potential climate keystone species.In Step #1, temporal trends of any relevant functional community property (e.g.community temperature index or carbon storage capacity) are quantified based on species' traits for the local community with and without the focal species.These temporal trends correspond to the community-level response to climate change.The difference in the slopes shows the contribution of the focal species.In Step #2, the degree of keystoneness is calculated for each species as the absolute value of the ratio between the contribution and the relative abundance of the focal species in the local community.In Step #3, the absolute contribution is contrasted with the absolute keystoneness.If the keystoneness can be fully explained by the contribution, climate keystone identification is not more useful than conserving species based on their dominance.In Step #4, the climate keystone species are defined as all those species whose keystoneness value exceeds the set threshold (signified with orange colour).In Step #5, the species are divided into groups based on their nonabsolute contribution sign (distribution on the left) and functional trait value in relation to the local community on average.

Validating the approach
Step #4 Selecting keystone species conditions, some resource and consumer species may contribute disproportionately to change in the average phenology of the food web.This can lead to changes in the structure of the food web as the spatiotemporal mismatches in existing resource-consumer interactions may force consumers to shift their diet (Deacy et al., 2017).
Given that we quantify the community-level response to climate change as the temporal trend in functional composition of the community, the community is observed at a fixed location over time but is not fixed in terms of the species it contains.
In order to identify climate keystone species, data of species' abundances or at minimum presence-absences in a spatiotemporal context and of the community property of interest are needed.The traits need to be selected carefully, for which we outline the caveats and recommendations in 'Limitations and future research perspectives' section.

| Step #1: Species' contributions
What: We quantify the species' influence (i.e.contribution) to the community-level response to climate change.
Why: The strength and sign of the species' contribution to the community-level response likely varies among species.Comparing scenarios with and without the focal species allows quantifying such variation in the contributions within the local community.
How: Various properties of communities' functionality, such as community temperature index or carbon storage capacity, can be used to measure the climate change response of a community (Devictor et al., 2012).Thereby, we calculate the community-weighted mean of the functional trait behind the community property of interest, such as temperature niche or carbon content of the species.
Thus, we account for the relative abundances of species within the community as well as the species-specific traits.To quantify the long-term community-level response to climate change, we calculate the temporal trend in the community property of interest for the local community using a linear model: We calculate species' contributions to this temporal trend using jackknifing, whereby we remove one species at a time from the dataset and re-estimate the temporal trend for the local community.We quantify the species-specific contributions to the temporal trend as the difference between the trend coefficient of the yearly change for the full community and for the community missing the focal species: A positive difference indicates that the focal species has contributed towards increasing the trend, while a negative difference indicates that the focal species has contributed towards decreasing the trend.
Why: To identify keystone species as objectively as possible, there is a need for a quantifiable measure of species' keystoneness.
With a quantitative measure of keystoneness, species can be ranked for keystone species selection.
How: We calculate the species keystoneness as the ratio between the species' contribution to the community-level response to climate change and the species' relative abundance within the local community: We consider absolute keystoneness values as we are interested in species' contributions to the community-level response regardless of their sign.Moreover, we use species' relative abundances rather than their absolute abundances within the local community to weight the contributions, because the ecological dominance of a particular species (approximated with relative abundance) may vary among communities even when the absolute number of individuals within the species remains constant.

| Step #3: Validating the approach
What: We confirm that in the local community of interest, there are species that contribute to the community-level climate change response more than expected by their relative abundance alone.
Why: To confirm that an approach for identifying keystone species is an efficient way to prioritize species for conservation, it must be established that some species contribute to the community's response disproportionately to their relative abundances.
Such validation has rarely been conducted in earlier studies (Lindenmayer & Westgate, 2020).Following the validation, if the species merely contribute by having large species-specific contributions to the community-level response, their keystoneness would be fully explained by the contribution of the species within the local community.
How: We test whether the species with high keystoneness values are the same species that have a large contribution in total across the individuals (high species' contribution values).That is, we test how the keystoneness depends on the contribution of the species within each local community.We use absolute values of the species' contributions and keystoneness to allow linear relationships between the variables.
Importantly, keystoneness is context(community)-dependent (Christianou & Ebenman, 2005).Therefore, we select the species that belong to the top quantile of the distribution of absolute keystoneness values in each studied local community separately instead of averaging keystoneness values for each species across large regions.

How: In
Step #2, we obtained the contribution of a species to the community-level response to climate change weighted by its relative abundance, similarly to the keystone species definition (Power et al., 1996).Now, we establish a relevant threshold and select all species whose absolute contribution exceeds the threshold as climate keystone species.

| Step #5: Categorizing keystone species
What: We divide the selected climate keystone species into categories based on their keystoneness direction (i.e.nonabsolute contribution sign) and climate role (relative trait value).
Why: Depending on the conservation target and the character-

| C A S E S TUDY: CLIMATE KE YS TONE B IRDS IN NORTH AMERIC A
To illustrate the use of the tool, we capitalize on the large-scale and long-term data of North American breeding and nonbreeding season bird communities (Meehan et al., 2019;Sauer et al., 2017).
In addition to repeating the five steps outlined above, we ask two questions on the functional and seasonal variation in keystone species: (1) Do other traits differ between climate keystone species of different climate roles (i.e.engineers versus bufferers)? and (2) Do contributions of climate keystone species differ between communities in breeding and nonbreeding seasons?In the bird communities, we expect some bird species to contribute to the community-level response more strongly than expected based on their relative abundances within the focal community.Due to the more intense climate change responses during winters (Lehikoinen et al., 2021), we also expect there to be more variation in the contributions to the community-level response in the nonbreeding season.In terms of functional differences, we expect that the groups of climate keystone species differ in their body size.Body size is positively associated with species' temperature tolerances and pace of life (Blueweiss et al., 1978;Peralta-Maraver & Rezende, 2021).Bartley et al., 2019;Devictor et al., 2008;Fridley et al., 2007;Gilman et al., 2010;Loreau & de Mazancourt, 2013;Neutel et al., 2002;Wallach et al., 2017).
For example, functionally diverse bird communities with more top predators (proxy of longer food chains, more complex network structure and higher degree of functional complementarity among species) have had smaller variation in their temperature niche over time (small variation indicating community stability) than bird communities with few top predators (Marjakangas et al., 2022).Generally, birds have important trophic roles for ecosystem functioning as they can be seed dispersers, pollinators, scavengers, insectivores or top predators (Sekercioglu et al., 2016;Tobias et al., 2022).Importantly, earlier studies have suggested trait-climate change and trait-ecosystem functioning links in bird communities, such that significant temperature variations seem to be responsible for both the Grinnellian and Eltonian aspects of functional homogenization (Gaüzère et al., 2015).
For the case study, we used monitoring time-series data of bird communities in breeding and nonbreeding seasons from 1966 to 2016 in North America (Meehan et al., 2019;Sauer et al., 2017; see Appendix S1 for details on case study data and methods).We included 546 species in the breeding season data and 609 species in the nonbreeding season data (species list in Appendix S2).To allow comparisons between breeding and nonbreeding seasons within the same standardized area and to account for stochasticity among survey units, we compiled survey units into grid cells of 5*5° across the study area (as in Lehikoinen et al., 2021).Moreover, the grid cell size roughly corresponds to average state/province size, at which resolution the conservation decision-making is commonly done.We totalled 60 grid cells in each season.We considered the grid cells as local communities of interest.
We applied the identification tool to the data.Following Step #1, we calculated the community temperature index (CTI; Devictor et al., 2008).CTI represents the weighted mean of species' temperature niches in the local community and is regularly used to measure community-level responses to climate change (Devictor et al., 2012;Lehikoinen et al., 2021;Tayleur et al., 2015).Temperature niches are indirectly relevant for ecosystem functioning under climate change because differences in temperature preferences can induce functional complementarity in the ecological community, which then translates into increased ecosystem functioning, such as higher seed production in pollination networks (Fründ et al., 2013).CTI can also be used to quantify thermal stability of communities under climate change (Marjakangas et al., 2022).For this, we obtained species' temperature indices (STI) from a recent study (Lehikoinen et al., 2021).
Using the season-specific STI values and species' yearly abundances in each grid cell, we calculated CTI for each survey unit within each grid cell in each year.To quantify the long-term community-level response to climate change, we calculated the temporal trend in CTI (hereafter, ∆CTI) for each local community separately using linear mixed models, where CTI was the response, year the predictor and survey unit the random variable ('lme4' R package; Bates et al., 2015).
First, we estimated the baseline of ∆CTI for each grid cell including all species in the estimation.To study how each species contributed to the baseline ∆CTI, we repeated the linear estimation while dropping each species one at a time from the dataset (jackknifing; Tayleur et al., 2016).Then, we calculated the difference between the baseline ∆CTI coefficient and ∆CTI coefficient of each repeated estimation where one species was dropped.This difference constituted the contribution of the focal species to the community-level response to climate change within each grid cell.A positive difference indicated that the focal species had contributed towards warming the community, while a negative difference indicated that the focal species had contributed towards cooling the community.

Following
Step #2, we calculated the keystoneness of each species in each grid cell using the ratio of the contribution and the relative abundance of the species.To obtain the species-specific relative abundances, the abundance of each species was first averaged across years and survey units in each grid cell and then divided by the sum of the averaged abundances across species within each grid cell.We did this to obtain a proxy for the relative abundances across all years and to ensure a consistent abundance calculation for all species.Following Step #3, we contrasted the species' absolute contribution and keystoneness using a linear regression, where keystoneness was the response and absolute contribution the predictor variable.Then, we assessed whether the climate keystone species provided an efficient conservation prioritization tool in this system by setting R 2 marginal < 0.05 as a cut-off for defining the efficiency.Following Step #4, we selected species belonging to the highest 1% quantile of absolute keystoneness value distribution in each grid cell.

Following
Step #5, we divided the selected climate keystone species into accentuating and dampening species based on their keystoneness direction values in each grid cell.Moreover, we divided the selected species into bufferers and engineers by calculating the difference in the species-specific STI value and the CTI of the local community including the focal species across all years.Following Fourcade et al. (2021), we obtained a relative STI value that can vary within species across communities.Species with a positive relative STI value were considered as engineers and the species with a negative relative STI value as bufferers in each grid cell.
To assess how engineer and bufferer species differ functionally, we compared their log10-transformed body sizes (as mass, g, Wilman et al., 2014) using a t-test.Moreover, we compared the keystoneness value distributions of climate keystone species between breeding and nonbreeding seasons.

Following
Step #3, we found that the keystoneness of species was only minimally attributed to the contribution of the species in both seasons (breeding season: R 2 marginal = 0.032, p < .001;nonbreeding season: R 2 marginal = 0.014, p < .001; Figure 2).Therefore, we explored keystoneness further.Following Step #4, in the breeding season bird communities, we identified 123 unique species as climate keystone species, of which 28 species were considered as climate keystone species in more than one grid cell.Similarly, in the nonbreeding season bird communities, we identified 124 unique species as climate keystone species, of which 27 species were considered as climate keystone species in more than one grid cell.N non-breeding = 30) based on relative STI values (Figure 2).We found that the engineer keystone species had significantly smaller body size than the bufferer keystone species in the breeding season, but not in the nonbreeding season (Figure 3, Appendix S1: Table S1).
Apart from the number of bufferer keystone species, breeding and nonbreeding season patterns in species' contributions did not largely differ.We also tested for differences in the population trend strengths between climate keystone and nonkeystone species and found that in general climate keystone species had stronger F I G U R E 2 Each row visualizes the results relating to Steps #3-#5 in identifying climate keystone species using bird community data in North America.For each step, results for breeding and nonbreeding seasons are illustrated separately (left and right panels, respectively).Season is indicated with grey symbols on the top-left corners of the panels.In Step #3, hexagon colour gradient indicates the number of observations (here, species in a grid cell) and the dashed line indicates the linear fit of the regression model.In Step #4, distributions correspond to the same observations, and for illustrative purposes, the selection of climate keystone species is exemplified with a 1% quantile threshold across all data instead of conducting the selection for each grid cell separately as was done in the actual analyses.The selected species across grid cells are illustrated with orange colour.In Step #5, the distribution of climate keystone species in the keystoneness direction (two plots at the top) and in the climate role (two plots at the bottom) is shown.population trends over time (for details, see Appendix S1: Section S4, Figure S2, Table S2).

| DISCUSS ION
Our definition of climate keystone species is formulated using the definition of Power et al. (1996) that considers a species as keystone if it has a large effect on the ecosystem in relation to its abundance.Alternative definitions of keystone species exist and a clear consensus is lacking (Cottee-Jones & Whittaker, 2012).For example, some authors focussed on the consequences of species removals and specifically on consumer (top predator) species (Cottee-Jones & Whittaker, 2012;Menge et al., 1994).We believe that a keystone species definition that accounts for the quantifiable contribution of a focal species to the community or ecosystem response relative to the other species in the community is more realistic in the contexts of climate change and ecosystem functioning.Indeed, removal of a single species would not lead to collapse of the entire ecological network in most ecosystems (Dunne & Williams, 2009).Such collapse is ever less likely under a press disturbance, such as climate change, that influences biodiversity gradually (Caro et al., 2022).Rather, the species-specific contributions to the community-level response to climate change should generally be small in absolute magnitude, although these absolute magnitudes may increase in species-poor communities.
Species' contributions to the community-level response may vary depending on their intrinsic traits as well as extrinsic stressors.For example, long-lived species may contribute more strongly as bufferers and short-lived species as engineer climate keystone species, because their temporal adaptation capacities likely differ.
Moreover, species that are strongly linked to other species (i.e.central in the interaction network), likely contribute strongly to the community-level response to climate change, because their increase and decrease have larger indirect effects on the community via other species (Dunne et al., 2002).Anthropogenic pressures beyond climate change can affect species' contribution to the community-level response to climate change, often producing cumulative and/or synergistic effects (Bowler et al., 2020).For example, land use change can restrict a species' ability to shift its range (Lawler et al., 2013) and make its contribution to the community response dampening even if the species exhibits a trait value, such as temperature niche, that would suggest an accentuating effect on the community response.
Moreover, populations of some species are artificially managed, for example with introductions and releases for hunting purposes, which may mask the imprint of climate change on them.

| Climate keystone species in North American bird communities
Using a long-term dataset of North American bird communities and focussing on their temperature niches, we identified several climate keystone species in 60 local communities across the continent.Although keystoneness is context-dependent, some species were identified as climate keystone species in multiple local communities.
For example, common redpoll (Acanthis flammea) was identified as a bufferer in four communities in the breeding season.We also found differences between the engineer and bufferer keystone species' F I G U R E 3 Body size (log10[g]) differences between groups of climate keystone birds.Groups are determined based on species' climate role: bufferer and engineer.Season is indicated with grey symbols on the top-left corners of the panels.Jittered circles illustrate the distribution of body mass values within each group of climate keystone species.The illustrated pairwise difference is statistically significant for the breeding, but not for the nonbreeding season (Appendix S1: Table S1).
body sizes in the breeding season, indicating that there are indeed particular regions of the trait space, and thereby ecological functions, that climate keystone species tend to occupy.Body size can be considered as a proxy trait for different life-history and resource use-related traits (Lim et al., 2020;McCain & King, 2014), such that engineer and bufferer birds may provide different ecosystem functions.In addition, we found seasonal differences in the climate role of keystone species such that there were proportionally more engineer species than bufferer species in the nonbreeding season, while in the breeding season the proportions of the climate roles were more even.This aligns with the fact that climate change effects on biodiversity are more intense during winter than breeding season, which has also been highlighted in other studies (Lehikoinen et al., 2021;Santangeli & Lehikoinen, 2017).Species' communities in winter (nonbreeding season) are under different selection pressures than communities in summer (breeding season), because the key demographic features in summer largely relate to reproduction and in winter to survival (Dingle, 1996;Pearce-Higgins et al., 2015).
In addition, many taxa in the most affected regions in the northern hemisphere migrate or have otherwise seasonal life cycles, making it important to assess the winter communities separately from the summer communities because of their only partially overlapping species compositions.Therefore, many relevant climate change-induced effects on biodiversity may go undetected unless nonbreeding communities are included in the studies.
In a broader context of climate change, our case study results provide novel information on whether the warm-dwelling species of lower latitudes or the cold-dwelling species of higher latitudes drive the observed community composition change.Earlier studies have mainly reported the overall changes in the communities' temperature niche, but have not specified which species drive the change (but see Princé & Zuckerberg, 2015;Tayleur et al., 2016).In particular, our case study indicates that the warm-and the cold-dwelling species contribute rather equally to the observed shift in the average temperature niche of the communities in the breeding season, while in the nonbreeding season the warm-dwelling (southern) species had stronger effects on the temperature niche of the communities.
Based on the keystoneness direction and climate role information and without additional population trend analyses, it seems that the abundances of warm-dwelling species have been increasing in the nonbreeding season over the study period (see Appendix S1: Figure S1 for details on the interpretation).This is consistent with the findings of recent studies showing northerly range shifts of many bird species during winter, including species like ducks that are ecologically and economically important (Meehan et al., 2021).

| Applications for conservation
We propose that the climate keystone species identification tool can be applied in conservation prioritization together with the existing measures for conserving the rare and endangered species.It can be used to identify species that preserve ecological communities and the associated ecosystem functioning most effectively (Figure 4).Before applying the tool, the delineation of the temporal and spatial extent at which the climate keystone species are identified should be made based on both the conservation policymaking needs and the ecological relevance (e.g.within a specific biome).Then, one must assess whether there are indeed climate keystone species in the community, or whether the functional property is driven by the F I G U R E 4 Decision tree for applying climate keystone species identification tool into conservation prioritization.The green boxes indicate questions that decision-makers should ask, and the orange boxes propose the kind of species to prioritize and the reasoning for the prioritization to achieve effective protection of ecosystem functioning.(Mammola et al., 2021;Mouillot et al., 2021) of different categories of climate keystone species can pinpoint functionally unique (groups of) species that should be conserved.On the contrary, if two groups of climate keystone species occupy the same trait space and thus likely provide the same ecosystem function, the most effective conservation should then target the group that can be protected with least effort.For example, if the trait spaces of climate keystone species with increasing or decreasing population trends in the local community do not differ, the most cost-efficient way to preserve the community via climate keystone species approach is to support the populations of increasing climate keystone species that do not require resource-intensive reintroductions or active protection.Although identification of climate keystone species would be the optimal strategy when considering the particular ecological community or ecosystem in isolation, it is best suited to complement other data supporting conservation decision-making.Finally, the time scale of the conservation target can affect the choice of climate keystone species for reaching conservation targets at different time scales.That is, in the short term, preserving species that maintain the current functional structure of the community can be considered a high conservation priority.Meanwhile, compositional turnover may be unavoidable due to global change, such that preserving species that support adaptation of community's functional properties may be a conservation priority in the long term (Dobrowski et al., 2021).
Concretely, climate keystone species identification could be applied to conservation for example by implementing the keystoneness as a metric in monitoring schemes, by using the identified climate keystone species as indicators when there are not enough resources to monitor all species, or by identifying priority species that should be target of more research, as they have a disproportionate impact on their local communities.For example, as Community Temperature Index (CTI) is being adopted by countries (e.g.Sweden and Finland) as an indicator of climate change impacts on ecological communities, identifying climate keystone species with CTI can help decision-making by showing which species influence the most the climate change responses of the vulnerable communities.This makes CTI a relevant functional property to consider in climate keystone identification as well.In general, if the trait that is being used to select climate keystone species is well established in the decisionmaking process or as an indicator, the climate keystone species identification tool can fit into existing protocols for conservation.

| Limitations and future research perspectives
A key assumption and potential limitation of the identification tool is that the selected functional trait must be relevant for climate change response and for ecosystem functioning of the taxon of climate change response can be challenging (Palacio et al., 2022), because it is not always straightforward to attribute the functional meaning of particular traits (Sobral, 2021;Violle et al., 2017).Moreover, dividing traits into response and effect (related to ecosystem functions) traits is context-dependent (Schleuning et al., 2020).We suggest that trait selection should be based on a thorough understanding of the natural history of the study system and the specific environmental context (Lavorel & Garnier, 2002;Schleuning et al., 2020).To this end, theoretical frameworks can provide clear criteria to characterize and prioritize traits within the ecological context of interest (Palacio et al., 2022).Despite the challenges, clear trait-climate links have been established, showing that populationlevel traits, even behavioural ones, are under the control of climate change (Cerini et al., 2023).As taxonomic groups generally differ in their functional and physiological traits, for example endotherms and ectotherms differ in their thermoregulation mechanisms, there is no universal climate response trait for all taxa, but the trait for climate keystone species identification needs to be selected taxonspecifically.We encourage future research to establish these traitresponse and trait-effect links for a wide range of taxa, particularly beyond primary producers that have been the main focus so far (Butt & Gallagher, 2018).
When applying the climate keystone identification tool, it is necessary to consider carefully the temporal and spatial scales at which the identification is done.The list of identified climate keystone species may vary depending on the temporal and spatial delineation of underlying data.In terms of the temporal scale, our case study illustrated clear seasonal differences in properties of North American climate keystone birds, showcasing the importance of temporal delineation of the climate keystone species identification.In terms of the spatial scale, climate keystone species identification should consider the relevance of the spatial unit from both ecological and conservation policymaking perspectives.In future, our approach may be adjusted to the temporal and spatial scale of interest by pooling survey units across national or biogeographical areas or focusing on finer scale communities or subsets of species within the communities.
The application of our tool may be limited by the availability of suitable data.Our case study of bird communities illustrates what is needed to identify the climate keystone species from empirical data.Beyond the local scale monitoring studies, the accumulation of (long-term) time-series data (e.g.Dornelas et al., 2018) allows the application of the tool at large spatial and taxonomical scales.Highquality data on species' traits relevant for climate change are needed to target particular ecosystem functions.
Our tool is developed to be applicable any taxonomic group and could be applied to both single taxonomic groups and to multitrophic communities and networks using different measures of functionality.However, we have only tested the tool with one taxonomic group and one trait in our case study; more case studies on different taxa are needed as the same traits and spatial scales will not be relevant for all taxa.The keystoneness values within a single taxon could be fine-tuned to account for species' centrality within the local food web, which allows giving more emphasis to species with many interactions and generally unifying different definitions of keystone species (Cottee-Jones & Whittaker, 2012).Few traits are relevant for ecosystem functioning and climate change response and measurable across trophic levels.Thus, the structure of the local interaction network could also be used instead of an average functional trait to quantify each species' contribution to the changes in a network structure measure, such as connectance or modularity, over time.Moreover, future studies could estimate species' contributions to community functionality (Step #1) with virtually any functional diversity measure (Mammola et al., 2021) beyond communityweighted trait means.For example, one could estimate the contribution of a species to the total volume of a probabilistic hypervolume (Mammola & Cardoso, 2020) or convex hull (Cornwell et al., 2006) representing the functional space of the community.However, the limitation of most summary measures across multiple traits is that the climate keystone species cannot be categorized based on their relative trait values (Step #5), which limits the derived conservation applications.
Finally, we did not test for the effect of the selected threshold in identifying climate keystone species.Therefore, we suggest that the climate keystone species identification tool can also be advanced by setting a more ecologically relevant threshold for selecting the climate keystone species.Indeed, the threshold could also be determined using simulations of tipping points and testing how many of the strongest contributing species in the local community need to be removed before a statistically different temporal trend in the community property of interest occurs.

| CON CLUS IONS
In this paper, we show a way to prioritize scarce resources for conserving ecological communities and the associated ecosystem functioning in the context of climate change.Identifying climate keystone species can represent a complementary tool to current conservation practices and allows better accounting for ecosystem functioning beyond the protection of single species (Miatta et al., 2021;Pollock et al., 2020).Therefore, we recommend using this tool as an addition within the conservation planning toolbox.
Climate change causes community composition turnover, which calls for dynamic approaches to conservation.For example, the climate keystone species identification tool could be applied together with dynamic conservation approaches, such as temporary protected areas (D'Aloia et al., 2019;Dobrowski et al., 2021), to ensure the protection of those areas and species that most effectively help reaching conservation goals of a set period of time.Moreover, the tool is highly applicable to ecosystems globally as it can be applied to any ecosystem function, taxonomic group or interaction network.
A similar keystone species identification could also be generalized to other anthropogenic stressors, including land use change, pollution, overexploitation or invasive species.We envision this as an efficient and cost-effective tool that will aid ecologists and conservation scientists in navigating complex and noisy biodiversity data and in identifying clear conservation and research priorities.This is a critical endeavour when facing climate and biodiversity crises whose dimensions are unmatched in human history.

ACK N O WLE D G E M ENTS
We thank Nicole Michel for the constructive feedback on the earlier versions of this manuscript.We also thank the volunteers for their efforts on BBS and CBC survey schemes.A.S., E.M. and A.L.
were funded by the Academy of Finland (projects 307909, 323527 and 329251).In addition, the research has been funded through the

CO N FLI C T O F I NTE R E S T S TATE M E NT
We declare no conflict of interest.

PE E R R E V I E W
The peer review history for this article is available at https:// www.webof scien ce.com/api/gatew ay/wos/peer-revie w/10.1111/ddi.13764.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data and code for conducting the case study are available on DRYAD: https://doi.org/10.5061/dryad.h9w0vt4mq.The raw bird survey data are available via the online platforms of the North American Breeding Bird Survey (BBS; http://www.pwrc.usgs.gov/)and the Audubon Society's Christmas Bird Count (CBC; https://www.audub on.org/conservati on/scien ce/chris tmas-bird-count).

Emma-Liina Marjakangas
https://orcid.org/0000-0002-5245-3779 Andrea Santangeli https://orcid.org/0000-0003-0273-1977 community − weighted mean of trait ∼ year contribution focal species = year slope full community − year slope full community−focal species keystoneness focal species = | contribution focal species ∕relative abundance focal species | Why: To effectively use scarce resources, clear guidelines for prioritizing species for conservation are needed (Arponen, 2012), such as by selecting keystone species.The concept of keystone species is commonly based on two axes: contribution and abundance (Cottee-Jones Figure S1).That is, a species can accentuate the community-level response either by having a larger trait value than the community on average and an increasing population trend (i.e.increased engineering, see definition below) or by having a smaller trait value than the community on average and a decreasing population trend (i.e.reduced buffering).Therefore, if we know the species' keystoneness direction and climate role, we can infer its population trend.How: First, we divide the species into accentuating and dampening species using their keystoneness direction values and assign them into categories such that species with a positive model coefficient value are considered as accentuating species and species with a negative model coefficient value are considered as dampening species.Second, we divide the selected climate keystone species into engineers and bufferers using species' relative trait values, such as temperature niches.That is, to define a continuum of engineer and bufferer species, we calculate the difference in the species-specific trait value and the average trait value of the local

Following
Step #5, we divided the climate keystone species into accentuating (N breeding = 79 of all identified climate keystone species across grid cells; N non-breeding = 44) and dampening (N breeding = 79; N non-breeding = 131) species based on keystoneness direction values.Moreover, we divided the climate keystone species into engineers (N breeding = 75; N non-breeding = 145) and bufferers (N breeding = 83; most contributing species (Step #2).If the keystoneness does provide added information, one must consider the details of the conservation target more carefully: what is the most appropriate spatial scale to identify the climate keystone species at and how many climate keystone species should be identified.Moreover, comparing trait spaces interest (MacLean & Beissinger, 2017).If the selected trait is not relevant for the taxon's climate change response, it is possible that other factors beyond climate change are the main cause for a species to have a strong contribution to the community-level change over time.These trait-climate change and trait-ecosystem function links are important given that the underlying goal of identifying climate keystone species is to conserve ecosystem functioning under climate change.The selection of relevant traits for communities' 2017-2018 Belmont Forum and BiodivERsA joint call for research proposals, under the BiodivScen ERA-Net COFUND programme, and with the funding organizations Academy of Finland (Helsinki: 326338) and the National Science Foundation (CLO, ICER-1927646).H.K. acknowledges funding from the Finnish Strategic Research Council's project IBC-Carbon (312559) and the EU Horizon project NaturaConnect (101060429).S.M. has received funding from the European Union's Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 882221 and the Biodiversa+2021 project DarCo.A.S. was funded through the European Union's Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 101027534 and the Jane and Aatos Erkko Foundation.