Transient demographic approaches can drastically expand the toolbox of coral reef science

Coral communities are threatened by an increasing plethora of abiotic and biotic disturbances. Preventing the ensuing loss of coral coverage and diversity calls for a mechanistic understanding of resilience across coral species and populations that is currently lacking in coral reef science. Assessments into the dynamics of coral populations typically focus on their long-term (i.e. asymptotic) characteristics, tacitly assuming stable environments in which populations can attain their long-term characteristics. Instead, we argue that greater focus is needed on investigating the transient (i.e. short-term) dynamics of coral populations to describe and predict their characteristics and trajectories within unstable environments. Applying transient demographic approaches to evaluating and forecasting the responses of coral populations to disturbance holds promise for expediting our capacity to predict and manage the resilience of coral populations, species, and communities.


Introduction
Coral populations worldwide are sensitive to a range of biotic and abiotic pressures (Riegl 2020), many of which will be exacerbated by future climatic change and human exploitation (Wolff et al. 2018;Riegl and Glynn 2020). Subsequently, the future status of coral reef ecosystems relies on the resilience of coral populations, specifically their capacity to endure, and recover from, increasingly frequent disturbance events (Holling 1973). Accordingly, a resilience-orientated focus is becoming commonplace in reef management (Mcleod et al. 2019(Mcleod et al. , 2021, necessitating an ability to predict the responses of coral populations to varying biotic and abiotic regimes (Lam et al. 2020). Yet, a limited understanding for the dynamics underpinning coral population performance and resilience still hinders our capacity to recognise and manage the climate induced reassembly and/or collapse of coral reef ecosystems (Bellwood et al. 2004;Edmunds and Riegl 2020).
Although limited, the application of demographic theory within coral research is not a novel concept (Edmunds et al. 2014). Structured population models offer insight into how the state of individuals (typically their size, age, or developmental stage ;Caswell 2001) arbitrates their survival and reproduction and how these fitness components in turn shape the overall responses of populations to local biotic and abiotic conditions (Box 1; Benton et al. 2006 Understanding how patterns in the survival, progression (i.e. growth), retrogression (i.e. negative growth, shrinkage; Salguero-Gómez and Casper 2010), and reproduction of individual colonies, influence population-level characteristics will improve effective management (Riegl et al. 2018;Brito-Millán et al. 2019a;Madin et al. 2020;Pisapia et al. 2020). Appropriately, state-structured demographic approaches are being increasingly utilised to quantify the mechanisms supporting or preventing the persistence and revival of disturbed coral communities (e.g. Kayal et al. 2018;Cant et al. 2021c). Accordingly, there is a growing appreciation for how the size structure of coral populations influences their dynamics and endurance Meesters 1998, 1999;Anderson and Pratchett 2014;Dietzel et al. 2020;Lachs et al. 2021), with recent appeals for efforts to be concentrated towards gauging the resilience of coral communities (Pisapia et al. 2020). However, we argue here that the contemporary application of statestructured demographic tools in evaluating the viability of coral populations, and their communities, does not reflect our need to anticipate their resilience.Prevailing knowledge of coral population dynamics centres around long-term (i.e. asymptotic) population characteristics (Box 2), with measures of long-term population growth rate (k) often used for projecting the viability of coral assemblages (e.g. Foster 2013, 2018;Edmunds 2015Edmunds , 2020Riegl et al. 2017;Kodera et al. 2020;Cant et al. 2021c;Shlesinger and van Woesik 2021). Indeed, the same is true across ecological theory, with our understanding of population dynamics focused on asymptotic population characteristics (Morozov et al. 2020). Quantifying the asymptotic characteristics of various populations presents a valuable comparative tool for exploring abiotic tolerances (Beissinger and Westphal 1998). Yet, for populations to realise their asymptotic potential, they require the persistence of stable conditions over extended timescales (Caswell 2001). In reality, though, seasonality exposes populations to periodic abiotic shifts, whilst predation, human exploitation, or storm and fire events generate recurrent and instantaneous disturbances. These short-term abiotic and biotic disturbances ensure that natural environments rarely exist at a stable equilibrium (Williams et al. 2011;Hastings et al. 2018). Consequently, evaluating the resilience and viability of populations requires an understanding for how the dynamics of populations manifest across timescales more representative of the disturbance cycles to which they are exposed (Hastings 2010). Here, we discuss why greater commitment towards quantifying the short-term, or transient, dynamics of coral populations is therefore required to enhance our capacity for anticipating the future resilience of coral reef ecosystems.

Transient versus asymptotic population dynamics
Structured population models characterise and define the trajectories of populations according to the individual-level allocation of resources towards survival, changes in size (or the state of interest), and reproduction. Populations can be structured according to the distribution of their individuals across some state variable, be that age (Leslie 1945), developmental stage (Lefkovitch 1965), size (Easterling et al. 2000), or a combination (Caswell et al. 2018 , and reproduction, and how they govern the characteristics of populations and their wider community interactions (Fig. 1). As corals mature and grow, they experience changes in their resource demands. The capacity of corals to continually meet these changing demands subsequently influences their ability to maintain their somatic condition and tolerate local biotic and abiotic stressors (Hughes 1984). The repeated survey of individual colonies, therefore, allows for documenting how the size of individuals regulates their survival, development, and reproductive contribution over time, and how this is shaped by changing environmental conditions (Boyce et al. 2006;Ehrlén et al. 2016). By condensing these temporal observations from across multiple tagged colonies, population ecologists can then explore how the individual-level vital rates subsequently underpin the characteristics of populations and define their capacity for tolerating various environments Crucially, although more time consuming, demographic approaches transcend the correlative techniques previously used for evaluating the viability of coral populations and communities, and for predicting their resilience to future climatic stressors (Edmunds et al. 2014;Edmunds and Riegl 2020). State-structured demographic approaches enable the quantification of the relationship between environmental conditions and population-level characteristics. Such an understanding is necessary if ecologists are to accurately predict the impact of recurrent disturbances and simulate populations under varying climate scenarios. Indeed, evaluating the vital rates and structure of numerous populations, and their association with abiotic regimes, has proven instrumental in calculating the extinction risk of highly threatened populations (Mace et al. 2008), and conservation of natural communities (Jongejans et al. 2008). Equally, combined with an explicit consideration for interactions between the dynamics of multiple populations, structured population analyses also facilitate the exploration of the mechanisms driving coexistence (Adler et al. 2010), and the transition of coral community composition amidst changing climatic conditions (Hughes 1996;Kayal et al. 2018;Cant et al. 2021c) mediated by selection gradients that emerge from the statespecific patterns of survival, development, and reproduction (Groenendael et al. 1988;van Tienderen 2000). In state-structured demography, patterns across these vital rates can subsequently be condensed into a population projection matrix (A), with each matrix element (a ij ) expressing the expected transition of individuals into state class i from state class j during the time interval t to t ? 1, or the per-capita a/sexual contributions of state j individuals into state class i during that same interval (Caswell 2001). This population matrix can then be used to project the size (N, number of individuals) and structure (n, number of individuals in each state) of a population: Fig. 1 Records of survival, size, and reproductive output, collected from tagged coral colonies over time, can enhance our understanding of population trajectories and, when collected across multiple populations, can be used to predict community reassembly under future climate scenarios. Here, subscript notation is used to reflect how the changing state of individuals corresponds with changes in individual-level characteristics. In this schematic the state of individuals refers to colony size and thus the growth (g) of individuals between size x and y is expressed as g yx Coral Reefs (2022) 41:885-896 887 Within a stable environment, the state structure of a population is expected to converge towards its stationary equilibrium, after which the state relative composition of the population will remain constant (Caswell 2001). At its stationary equilibrium, therefore, the size of a population is expected to change at a constant rate through time, termed its asymptotic population growth rate (k; Caswell 2001). This value of k, for any state-structured population, is equal to the dominant eigenvalue of its corresponding population

Box 2 The status of demographic tools within coral research
Conducting a review of the relevant ecological literature, we examined the advancement of state-structured population approaches within coral research. Specifically, we searched the SCOPUS database for the term coral* AND demograph* across the titles, abstracts, and keywords of peer-reviewed research literature published between 1960 and 2020, inclusive. This search initially returned 781 studies, from which we retained all studies that explicitly measured or quantified state-structured patterns in the survival, size transitions (including shrinkage via partial mortality or fission), and/or reproduction (e.g., fecundity) of coral colonies. We also retained studies measuring larval settlement rates, as the recruitment of early life stage individuals is a key demographic aspect within coral communities (Adjeroud et al. 2017). Following this refinement, we retained a total of 145 studies (* 19% of our original search), which we then categorised according to whether they involved the presentation of (i) asymptotic dynamics/simulations, (ii) transient characteristics, or (iii) an assessment of vital rate patterns only (see Supplementary materials) Following the seminal work of (Hughes and Jackson 1980), the application of state-structured demographic assessments has increased within coral research (Fig. 2). Most of this research (63%; 91 studies) reports no measure of population-level characteristics, focusing instead on evaluating patterns across one or more of the vital rates of survival, growth, retrogression, or reproduction. Meanwhile, 50 studies (34%) focus on the analysis of deterministic asymptotic (i.e., long-term) population characteristics (e.g., population growth rate, k), with nine studies (6%) also reporting measures of stochastic population growth rate (k s ; Lewontin and Cohen 1969). Estimates of k s describe the average periodic change in the size of a population over a series of successive time intervals (Engen and Saether 1998), and thus provide a greater consideration of environmental variation than deterministic k. However, with estimates of k s often derived from a projected series of discrete population growth rates with initial estimates of population size omitted, k s is still fundamentally a long-term measure of population viability (Ellner and Rees 2007). Crucially, across our search, just three studies (2%) explored the transient (or short-term) characteristics of coral populations (see Hughes and Tanner 2000;Linares et al. 2007;Bruno et al. 2011) Fig. 3 At stationary equilibrium, populations are expected to change size at a constant rate, termed their asymptotic growth rate (k). However, to display asymptotic characteristics, populations need a stable state structure and so require time at equilibrium to enable the development of this stable structure. In the absence of a stable state structure, populations will display varying growth rates as they converge towards asymptotic expectations, even within equilibrate environments. Note that, although in this schematic the populations growth rate is increasing towards its asymptotic trajectory, estimates of k can also reflect asymptotic decline Fig. 2 The cumulative number of studies that explicitly explore the state-structured demographics of coral populations as identified by our search for the term coral* AND demograph* across the peerreviewed literature within the SCOPUS database. The colour scale differentiates between studies that focus on only investigating patterns in colony survival, growth, and/or reproduction, and those that present estimates of either deterministic population growth (k), stochastic asymptotic characteristics (k s ), or transient population characteristics Alternatively, the measures of first-step attenuation and maximal attenuation can be used to describe the initial, and maximum decline in population growth rate during transient conditions. Meanwhile, damping ratio estimates describe the time required for a perturbed population to reobtain a stable equilibrium. Accordingly, by revealing the capacity for populations to grow (compensation), avoid declines (resistance), and/or recover following a disturbance event, these transient metrics present ecologists with measures of population resilience

Box 3 Measuring transient population characteristics
Within natural environments, abiotic and biotic perturbations generate constant shifts within the state structure of populations ensuring they persist within a transient state, rather than converging towards a stable equilibrium (Hastings 2004). Within this transient state, the growth rate of populations can differ dramatically from their asymptotic trajectories, with populations undergoing amplification (increases in growth rate) or attenuation (declines in growth rate; Fig. 4). These transient properties subsequently influence the viability of populations in the longer-term, underpinning their ability to compensate (i.e., benefit), resist, and recover following disturbances (Fig. 4). Accordingly, it is necessary to explore the transient characteristics of populations to gain insights into their responses to disturbance, and therefore better understand their resilience (Ezard et al. 2010;Capdevila et al. 2020) Evaluating the transient dynamics of populations, and therefore their demographic resilience, is made possible through the calculation of transient metrics readily accessible from population projection matrices (see Stott et al. 2011;Capdevila et al. 2020 for a detailed review). These measures include population reactivity (q) and first-timestep attenuation (q) which describe the increase or decline in a populations growth rate within one time interval of a perturbation, relative to k (Stott et al. 2011). Equally, the measures of maximal amplification (q max ) and maximal attenuation (q max ) can be used to evaluate the maximum amplification or attenuation expected in the growth rate of populations, relative to k (Townley et al. 2007;Townley and Hodgson 2008). The damping ratio (q) of a population defines the rate at which a population will converge back to a stable equilibrium following a perturbation to its stable structure (Caswell 2001 matrix A and exists on a non-negative scale reflecting the proportional growth (k [ 1), decline (k \ 1), or demographic equilibrium (k = 1) of a population over time. For their stable structure to manifest, however, populations typically require time in a constant environment to allow the state distribution of their individuals to converge towards its equilibrium state without displacement by local disturbances (Fig. 3; Bierzychudek 1999). Yet, populations are rarely exposed to stable conditions or equilibria (Hastings 2001(Hastings , 2004Williams et al. 2011), and the vital rate patterns used in constructing population projection matrices are themselves not consistent and contingent on the environmental conditions in which they are observed (Benton and Grant 1996;Boyce et al. 2006). Moreover, as populations change in size their dynamics are subject to density-dependant regulation Lande et al. 2003). The dynamics of populations are, therefore, expected to vary over time as environmental conditions change (Tuljapurkar and Orzack 1980;Tuljapurkar 1989). The perpetual instability generated by regular disturbances ensures that populations rarely attain their asymptotic trajectories (Hastings 2001;Hastings et al. 2018). Natural environments persist within a non-equilibrate condition maintaining populations within a transient state during which their trajectories can radically differ from asymptotic expectations (Stott et al. 2010(Stott et al. , 2011Hastings et al. 2018). Anticipating the future condition and endurance of populations thus requires an awareness for their transient characteristics (Ezard et al. 2010). Transient shifts in the state structure of populations can elevate (amplification) or diminish (attenuation) their Box 4 Transient dynamics in Pocillopora aliciae: an empirical example Here, we illustrate how the asymptotic and transient dynamics of coral populations can indicate very different realities regarding their continued viability, whilst also demonstrating the role of transient demographic theory in understanding coral population performance within variable environments. Pocillopora aliciae (Fig. 5) (Cant et al. 2021c). Our IPM (Eq. 2) describes how the size structure of the P. aliciae population in year t?1, n ðz 0 ;tþ1Þ , is a product of its state structure in year t, n (z, t) , and the state mediated patterns of survival (r), growth (c, including shrinkage and stasis), fragmentation (j, comprising its probability and both the number and size of any remnants produced), and recruitment (f, the number and size of new recruits). Within this IPM, patterns in individual survival, growth, fragmentation, and recruitment, are condensed into a probability density kernel (K; Eq. 3) outlining the transition of individuals across a continuous size spectrum (z to z'; Fig. 5), and thus analogous to the discretised population matrix A in Eq. 1 n ðz 0 ;tþ1Þ ¼ R r ðzÞ c ðz 0 ;zÞ j ðzÞ f ðzÞ n ðz;tÞ dz (2) ¼ R K ðz 0 ;zÞ n ðz;tÞ (3) From our IPM we then obtained estimates of asymptotic population growth rate (k), alongside each of the transient metrics of reactivity (q), first-timestep attenuation (q), maximal amplification (q max ), and maximal attenuation (q max ; Box 3) In the Solitary Islands Pocillopora aliciae appears to be in a state of rapid decline (Fig. 6a). Estimates of k we obtained for the population correspond with negative population growth (k \ 1; k: 0.704 [95% CI: 0.698, 0.710]), with the population projected to disappear from the region within *15 years (Fig. 6a). This trajectory is possibly a consequence of extensive bleaching observed throughout the Solitary Islands during 2015-16 (Hughes et al. 2017;Kim et al. 2019). However, subtropical coral populations routinely possess low rates of long-term population growth, particularly in comparison to their tropical counterparts (e.g. Cant et al. 2021a, b). Yet, despite reduced long-term performance estimates inferring their diminished future viability, coral populations continue to endure within subtropical environments (Mizerek et al. 2021) Exploring the transient dynamics of the P. aliciae population we reveal characteristics associated with the elevated tolerance of recurrent disturbance regimes, with the population demonstrating a capacity for demographic compensation (i.e., an increase in population growth following a disturbance to its structural composition; Fig. 6b). The population exhibits a reactivity (q) of 2.092 [1.985, 2.198] reflecting the doubling of population growth rate within one year of a disturbance, relative to k. During its transient phase, the growth rate of the population is then expected to peak (q max ) at 2.226 [2.111, 2.342] times k, before settling at approximately twice that expected by longterm growth estimates (Fig. 6b). However, unlike asymptotic characteristics, transient dynamics are contingent on initial population structure (Stott et al. 2011), and changing the structure of the P. aliciae population does influence its transient trajectory (Fig. 6b). An increased bias towards larger sized individuals ([80% of individuals larger than median size) elevates its reactivity ( . Although, this structural shift represents a radical bias and results in only a 20% reduction in population growth, relative to k. With subtropical coral populations, such as P. aliciae, exposed to highly seasonal environments compared to tropical reef populations (Sommer et al. 2014), the ability to enhance population growth following disturbances may represent a competitive advantage (Cant et al. 2021a, b). Accordingly, it is necessary we move beyond long-term measures of population performance when evaluating the viability of coral communities growth rates relative to asymptotic projections (Townley et al. 2007). Equally, for populations existing in a transient state, there is a natural tendency for their state structure to converge towards a stable distribution (Caswell 2001). Intuitively, with population resilience comprised of the attributes of resistance (the ability to withstand disturbances), compensation (the ability to benefit from disturbances) and recovery (the speed and ability to reattain stationary equilibria), the transient dynamics of populations therefore represent quantifiable measures of these features of resilience (Box 3; Hodgson et al. 2015;Capdevila et al. 2020).

Fig. 5
Describing how size mediated patterns in survival, growth, fragmentation, and recruitment, determine the transition of individual colonies during the time interval t to t ? 1, we used an IPM to evaluate the condition and viability of a Pocillopora aliciae population. Repeated surveys of individually tagged colonies were used to document the survival, growth, fragmentation, and recruitment characteristics of individuals colonies, and how these patterns are influenced by colony size. By condensing these individual-level patterns into a probability density kernel our IPM subsequently translates the transition of P. aliciae individuals across a continuous size spectrum into measures of both asymptotic and transient population characteristics Fig. 6 The asymptotic and transient dynamics of Pocillopora aliciae within the Solitary Islands forecast very different outlooks for this subtropical population. (a) Estimates of long-term population growth correspond with a rapid temporal decline in population size; a trajectory that persists irrespective of population structure, as illustrated by comparisons between forecasts involving a population in which individuals are distributed evenly across the size spectra (black), a population with a structural bias towards larger sized individuals (red; [80% of individuals larger than median size), and a population with a structural bias towards smaller sized individuals (blue; [80% of individuals smaller than median size). (b) Meanwhile, the transient characteristics of the population indicate it to possess the capacity for elevating its growth rate, relative to long-term expectations (k, dotted line), following a disturbance. However, the magnitude of this increase is dependent on initial population structure, with the increased prevalence of smaller-sized colonies reducing population reactivity (q), and driving population attenuation (diminished growth rate relative to k). Across both panels shaded regions correspond with the confidence bounds associated with each projection Transient dynamics and the resilience of coral communities Transient demographic approaches present an effective tool for explicitly evaluating the resilience of natural populations (Capdevila et al. 2020). Indeed, such approaches have proven instrumental in our understanding of the establishment of populations within new and variable environments (Iles et al. 2016;McDonald et al. 2016McDonald et al. , 2017Jelbert et al. 2019). Exploring the transient dynamics of coral populations can thus enhance our understanding regarding the future diversity, condition, and resilience, of coral communities. Globally, coral communities are being exposed to increasingly recurrent thermal stress events, at frequencies soon expected to exceed the recovery time required by most coral species (Hughes et al. 2018;Sully et al. 2019). Presently, considerable value is placed on classifying coral species according to their morphological and functional traits, with the subsequent categorisation of coral taxa offering proxies for the vulnerability of various coral communities to said recurrent stress events (Darling et al. 2012(Darling et al. , 2019. Quantifying the relative affinities of different coral populations for demographic recovery presents the opportunity for justifying these assumptions; helping to distinguish the species and populations most vulnerable and or resilient to projected recurrent disturbance regimes. Equally, the amplification and attenuation characteristics of populations simultaneously reflect their stability following a disturbance. Distinguishing between the differential attributes of compensation and resistance in different populations is important for managing the future condition of biological communities (Capdevila et al. 2020). Assessments of coral communities following repeated disturbances have demonstrated divergent shifts in species compositions (Adjeroud et al. 2009;Pratchett et al. 2011Pratchett et al. , 2020. With knowledge of the capacity for various coral populations to undergo demographic amplification or for avoiding demographic attenuation, we can begin to forecast and anticipate the reassembly of coral communities, thereby improving the effectiveness of adaptive reef management. Indeed, Sandin and McNamara (2012) and Brito-Millán et al. (2019b) demonstrate how the transient characteristics of coral populations determine the resilience of their wider reef communities to algal phase shifts. Importantly, these contributions to our understanding of transient dynamics within coral environments advocate for a consideration of spatial interactions within assessments of coral population dynamics. Spatial interactions between individual corals impacts upon their fitness characteristics (Buss and Jackson 1979;Idjadi and Karlson 2007;McWilliam et al. 2018), yet the exact role these spatial interactions play in defining the performance of coral populations remains unresolved (Alvarez-Noriega et al. 2018). Ultimately, assessing the combined influence of state-structured demography and spatial interactions (e.g. Adler et al. 2010;Chu and Adler 2015) upon the performance, and resilience, of coral populations represents an important and worthwhile endeavour.
Alternatively, a disturbance does not necessarily imply a negative impact, merely a change to a system (Pickett and White 1985). Disturbances to the structure of populations can be brought about by the cessation of a continuous pressure (e.g. Götz et al. 2008), changes in resource availability (e.g. Goldstein and Steiner 2020), or their introduction into a novel environment (Jelbert et al. 2019). With the ability of natural populations for undergoing demographic compensation underpinning their capacity for exploiting changes in their local environment (Jelbert et al. 2019), the transient dynamics of populations can offer insight into their invasive potential (Iles et al. 2016). Demographic compensation mediates the persistence of coral populations established along coastlines characterised by increased abiotic variability (Cant et al. 2021a, b). Predicting and quantifying demographic compensation within coral populations will, therefore, aid in forecasting the winners and losers in reef communities targeted by conservation and restoration initiatives, or those undergoing climate induced range shifts. To illustrate this potential, we provide an example evaluating the performance of a subtropical coral population in Eastern Australia (Box 4).
Complexities in the modelling approaches used to explore the dynamics of natural populations have resulted in these techniques remaining largely overlooked within coral research (Edmunds et al. 2014). Indeed, parametrising the demographic models needed to quantify population characteristics requires considerable amounts of data (Ellner et al. 2002). Although, despite the data demanding nature of state-structured demographic models, the collection of the necessary data regarding the survival, transitions in size, fragmentation, and recruitment, of individual coral colonies is possible alongside current reef monitoring efforts . Furthermore, new techniques for analysing the demographic characteristics of populations structured by a continuous state variable like size (Integral Projection Models; Easterling et al. 2000) have reduced the data demands of demographic assessments (Ramula et al. 2009). As a result, IPM frameworks are growing in popularity throughout coral research (see Bruno et al. 2011;Madin et al. 2012;Zychaluk et al. 2012;Elahi et al. 2016;Kayal et al. 2018;Precoda et al. 2018;Lord et al. 2020;Cant et al. 2021c;Carlot et al. 2021). Finally, studying the dynamics, particularly the transient characteristics, of coral populations is made increasingly accessible by the development of demographic analysis packages such as popdemo (Stott et al. 2012) within open-source programs like R (R Core Team 2019).
Forecasting the resilience of coral populations presents a major challenge for the future conservation of various reef communities (Lam et al. 2020). Currently, we lack the mechanistic insight into the resilience and response of coral populations to disturbance necessary for predicting the future reassembly and persistence of coral communities given increasingly recurrent disturbances. It is not our intention to dismiss previous efforts to evaluate the dynamics of coral populations using asymptotic characteristics. The asymptotic characteristics of populations present a useful tool in exploring the relative tolerances of differing populations (Beissinger and Westphal 1998) and have proven valuable in the management of threatened populations (Johnson et al. 2010). We have illustrated, however, that for evaluating and forecasting the responses of coral assemblages to disturbance and changing environments, asymptotic characteristics do not represent measures of population resilience (Hastings et al. 2018). Instead, we advocate for greater a focus towards investigating the transient dynamics of coral populations that more accurately reflect their dynamics within natural environments (Ezard et al. 2010). Evaluating coral reef community resilience using the framework of transient demographic theory will provide a new effective lens on the temporal and spatial nature of coral reef resilience, a view that will transcend current approaches of conducting single reef assessments to determine resilience (Maynard et al. 2015;McLeod et al. 2021).
Acknowledgements The authors would like to thank D. Ceccarelli for her comments on early drafts of this manuscript. This work was supported by a NERC DTP scholarship to JC and a NERC Independent Research Fellowship (NE/M018458/1) to RS-G.

Declarations
Conflict of interest On behalf of all authors, the corresponding author states that there is no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/.