The myth of resilience: Necessary non-resilience and the role of variable choice in ecosystem recovery

Resilience, the ability to resist or recover from disturbance, is ubiquitous in ecology but defined and measured in different ways. The evaluation of resilience depends on decisions made by the investigator(s), including the variables measured. Here we highlight an under-appreciated observation: there is no canonical definition of overall resilience and such a definition may be unattainable. Therefore, we make four key points. First, we highlight and categorize the diverse variables used to measure ecological resilience and place them in a conceptual model. Second, we argue that different relevant variables often respond very differently to disturbance and prove that no system can be completely resilient to a press disturbance (‘Necessary Non-resilience’). Third, we demonstrate with four examples how categorization of diverse resilience variables and a conceptual model can stimulate new research questions. Fourth, we apply our framework to four empirical case studies to demonstrate the biological relevance of such new directions. Overall, we argue that advancing resilience ecology will require a deeper consideration of variable choice, how different resilience variables interact, the inevitable failure of resilience in some variables, and how these ideas can foster new, general research directions.

Ecological resilience is an important, widely-used, and occasionally divisive concept that provides a framework for discussing how ecological systems respond to disturbance (Capdevila et al. 2021; Mori 2016; Thorén 2014). Broadly, resilience is defined in terms of an ecological system’s long-term ability to maintain a particular state, often with some desirable characteristics or functions, in the face of a disturbance (Holling 1973). As one widely-cited example, Hodgson et al. (Hodgson et al. 2015) define resilience as the property of a system to resist change and maintain the current state despite a disturbance (resistance) or its capacity to recover to that state after a disturbance (recovery). Together, these two properties imply that a more resilient system will be closer to its pre-disturbance state following a disturbance when compared to a less resilient system. This definition applies to both transient pulse disturbances (e.g., a hurricane, fire, vegetation removal) as well as steady, long-term press disturbances (e.g., increased mean temperature, the establishment of an invasive species). Note that we never consider the unlikely case of systems that are 100% resistant in every variable to a given disturbance. Thus, resistance and recovery together help define resilience. However, a general definition of resilience is not itself sufficient to classify a particular ecological system as resilient. Operationalizing resilience requires an additional key step – choosing one or several relevant variables to measure. This decision about which variables are relevant to determining resilience must be made by the investigator(s). For example, in the same forest ecosystem following a fire event, one investigator might reasonably choose to focus on the long-term recovery of tree species richness as an indicator of resilience, while another might evaluate overall biomass, while yet another might focus on primary productivity. Thus, the assessment of and conclusions regarding resilience may differ markedly depending on the perspective and selected metrics. In the resilience literature, investigators often interpret the resilience of their chosen variable as the resilience of the system as a whole (van Der Loop et al. 2023; Hensel et al. 2021; Li & Wang 2023), implicitly or explicitly. This interpretation would be reasonable if all meaningful variables could be expected to lead to qualitatively similar conclusions regarding the resilience of a system. Unfortunately, there are many counterexamples to this assumption throughout the literature (Orr et al. 2024). For example, species richness of salt marsh plants was not resilient to experimental extirpation (an artificial disturbance) while productivity and biomass were, up to a point (Davies et al. 2012). Similarly, plant population abundance can be considered resilient in the face of global change due to shifts in plant phenology. However, plant phenology displays much less resilience in this case (Cleland et al. 2007). Even very closely linked variables can respond differently; for example, community biomass could be constant, due to increasing abundance, while mean body size declines (Martins et al. 2023). In the case of invasive species ecology, general ecological resilience has recently been described with the variable choices of invasive plant biomass (van Der Loop et al. 2023), native plant cover (Hensel et al. 2021), or ecosystem services provided by native and invasive plants (Nyssen et al. 2024), which can all respond very differently. For example, invasive plants may capably provide ecosystem services (a common occurrence Nyssen et al. 2024). These differing responses highlight the central role of variable choice in shaping conclusions about resilience. The current tendency to neglect the importance of variable choice risks mischaracterizing resilience, generating conflicting interpretations among stakeholders, producing inaccurate predictions of system responses to future disturbances, and inhibiting integration across subfields of resilience ecology. Failure to acknowledge variable diversity and choice can lead to poor predictions of resilience to future disturbances. As a classic example, large declines in population abundance can lead to decreased genetic diversity through a bottleneck effect that persists after population abundance has recovered (Hoelzel et al. 2024; Keller et al. 2001). An investigator only considering resilience of the population abundance variable may conclude the system is quite resilient, not realizing that degraded genetic diversity renders the population much more vulnerable to a future disturbance. Additionally, while there have been increasing calls for integration across subfields of resilience (Dakos & Kéfi 2022; Reed et al. 2024; Thorogood et al. 2023), these calls aim for the most generally applicable ways to quantify resilience in a single variable; but we argue that ecological resilience involves multiple interacting, potentially conflicting variables so that general integration across subfields will advance more successfully if investigators embrace the difficult realities of variable diversity. Overall, we aim to highlight the value of considering multiple, interacting resilience variables and offer a path for advancing resilience science that allows for a more nuanced and mechanistic understanding of ecosystem resilience. Recognizing the importance of variable choice, we provide four interrelated points about resilience variables. First, there is a far greater diversity of variables that could be used to measure resilience than are typically evaluated in any one study, and we provide a framework to categorize these variables into meaningful groups along with a conceptual model of interactions among groups. Second, conclusions regarding resilience will nearly always depend on the particular variables chosen to measure resilience. We discuss the high probability that different important variables will differ widely in their resilience to a single disturbance and prove more formally that no system can be resilient in all possible variables to a press disturbance. Third, our framework can help to generate research questions and testable hypotheses about resilience variables, and we propose four examples we consider especially compelling for future work. Finally, the application of this thinking to case studies from empirical systems demonstrates how it can lead to novel insights and research directions in ecological resilience. Categorization of diverse resilience variables A great diversity of variables are used to measure ecological resilience (Kéfi et al. 2019). To aid general progress, we must first acknowledge and organize this diversity of variables. We do so by dividing commonly-measured variables into categories to aid the formation of general research questions and hypotheses about variable categories. Our chosen categories contain this diversity of variables and we provide key, illustrative examples in each category (see Appendix for a larger table containing all variables compiled by Kéfi et al. 2019 as well as several additions in the same categorization as Table 1). We divide variables into Amounts, Characteristics, and Functions (see Table 1). Amounts concern the number, biomass, biovolume, etc. of living individuals. Characteristics capture many properties of organisms that are likely to interact with their fitness, positively or negatively, including phenotypes, genotypes, vital rates, age, etc. Functions are properties that do not have a clear relationship with fitness but are of ecological interest for other reasons, e.g., aesthetic value (Tribot et al. 2018). All three of these categories apply across the classic ecological scales that we consider; an individual organism, a population of individuals of the same species in a defined area, and a community of multiple species in a defined area (Odum & Barrett 1971). We note that different scales could also be considered including molecules, meta-populations, etc.; for the sake of simplicity, we focus on the three classical scales but our core arguments apply more broadly to other scales. Individual-scale variables are measured on a single individual, potentially a marked or captive individual over time, while population-scale variables are measured on multiple individuals of a species and community-scale variables are measured on populations of multiple species. Further, variables at the population and community scales are further subdivided into full distributions (the frequency of different values observed across individuals or across species) vs. metrics (different aggregations of these population- or community-level distributions). Table 1. Categorization of resilience variables across scales. We consider resilience variables across the individual, population, and community scales. At the population and community scales, resilience variables may be considered as the full distribution across individuals or species or as a single, aggregate metric (we use numbers 1., 2., 3., etc. to link distributions to the corresponding aggregate metrics within a scale, e.g., the biomass metric is the sum of the distribution of body sizes); further, both distributions and aggregations may or may not be normalized to be independent of abundance. At all scales, we divide variables into three categories: Amounts, Characteristics, and Functions. We categorize a set of example variables that are far from exhaustive (see Table A1 for a larger set of examples). We note that prior studies have also measured resilience using abiotic variables, which are not included here. | Scale | Distribution or Metric | Amount | Characteristic | Function | | Individual | Not applicable | 1. Body size, 2. Alive | Phenotype (Behavior, physiology, microbiome), Genotype, Age, Hazard, Lifespan, Fecundity, Growth Rate, Infection status | Aesthetic appearance | | Population | Distribution | 1. Body sizes | 1. Genotypes,2. Individual ranges,3. Vital rates, 4. Ages,5. Sexes | 1. Energy flows, 2. Nutrient fluxes | | Metric | 1. Biomass, mean body size, etc.2. Abundance/cover | 1. Genetic diversity,2. Species range/Connectivity3. Population growth rate, 4. Median age,5. Sex ratio | 1. Energy flow, 2. Nutrient flux | | | Community | Distribution | 1. Biomasses,2. Abundances/covers, | 1. Interaction strengths,2. Growth rates | 1. Energy flows, 2. Guild memberships,3. Nutrient fluxes, 4. Pest consumptions | | Metric | 1. Biomass,2. Abundance/evenness/species richness/cover | 1. Ecological network stability,2. Total growth rate | 1. Energy flow, 2. Number of guilds, 3. Nutrient flux,4. Pest consumption | To consider the relationships among these diverse resilience variables across scales, we provide a conceptual model of their interactions and response(s) to disturbances. Other conceptual models could be devised but should contain many similar core elements in terms of causal relationships, scale relationships, evolution, and plasticity; explicit conceptual models, such as ours, will aid investigators in making their case for why they expect certain resilience variables and relationships among them to be important. We ground our conceptual model by considering disturbances as acting on individual organisms (green arrows from disturbance to individuals in Fig. 1). Disturbances may affect the phenotype of an organism via plastic change (Reed et al. 2011), such as behavioral shifts (Stephenson & Adelman 2022). Phenotypes are also determined by the individual’s genotype and affect vital rates. Alternatively, or in addition to modifying phenotypes, disturbances may directly affect the vital rates of an organism by killing it (Palkovacs et al. 2012) and altering its lifespan or indirectly modifying its fecundity, e.g., by altering resource availability. We define individual-level variables as those variables that can be measured for a single individual, such as body mass, behavior, genotype, lifespan, etc. (see Table 1, first row). Disturbances are just one exogenous factor acting on an individual’s phenotypes and vital rates, however; the abiotic environment and ecological community may also impact individual phenotypes and vital rates. We then aggregate across individuals up to the population scale . We define population-level variables as those sampled by measuring multiple individuals of a single species in a defined area. We consider population distributions of variables and population metrics that often condense information from population distributions (see Table 1 and black arrow in Fig. 1). Distributions contain the full information of how many individuals have a certain phenotype, home range, vital rate, genotype, etc. The distribution of phenotypes influences, for example, the spatial distribution of a population and its vital rates. The relationship between the distribution of vital rates and the distribution of alleles affects the distribution of alleles in the next generation via selection. Immigration can also act as an exogenous factor, altering the distribution of alleles. While ecologists sometimes use the full information of population distributions (as far as it is accessible; Newman et al. 2020), we also often collapse this information into metrics, commonly a total aggregating across a distribution. For example, we may care about the total count of females, the total function of a population, or its total range. Some of these population-level totals have classical relationships with each other. For example, population growth rate (which may also depend on immigration) determines the change in abundance while energy flow determines the change in biomass (neglecting changes in temperature or caloric density). We can then aggregate across populations up to the community scale . In a community, we conceptualize species interaction networks as the impact of a population of one species on individuals of another species through plastic effects (dotted green arrows in Fig. 1) or effects on vital rates (dashed green arrows). Similar to population variables, ecologists often consider the full distribution of a community variable, such as the distribution of species abundances (Supp et al. 2012). At other times, ecologists aggregate across distributions to obtain single metrics, such as total abundance or species evenness. We note that this conceptual model in Fig. 1 illustratively captures a large number of resilience variables and their common interactions; for parsimony, we do not attempt to capture all possible variables and interactions (e.g., those involving maternal effects). The depiction of scales in Figure 1 further reflects the importance of investigator choice. For example, the microbiome can be depicted most simply as an individual-scale characteristic or more fully as members of the community which may be nested within an individual, depending on the focus of the investigator. Similarly, the defined spatial extent of a population will alter the importance of immigration. Regardless of the investigator’s choices on these matters, Figure 1 still emphasizes the importance of investigator choice of resilience variable, or set of variables, among the many, measurable, likely interacting resilience variables across scales and categories. Differing resilience responses and Necessary Non-resilience Examination of Table 1 suggests three observations. First, it is unlikely/impractical that a complete set of all possible variables could be measured in a real ecological system. Second, as a result, investigators will generally need to choose a subset of all possible variables to measure. Third, there does not appear to be any way to select an obviously “correct” variable or canonical set of variables to assess ecological resilience, instead of other sets of variables. As an example of contrasting priorities, public engagement efforts often emphasize the importance of the survival of a specific individual while managers may emphasize larger scales, such as the population. This difference in priorities was exemplified by expressions of disapproval from the public after the Tennessee Wildlife Resources Agency euthanized 13 named, black bear cubs rather than release them back into the wild due to the spread of an infection (Krueger n.d.). The agency’s mandate is to promote population-level resilience where the public was concerned with individual-level resilience. Further, cultures often differ in their priorities for and questions regarding the natural world (e.g., Native American vs. European American; Medin & Bang 2014). Even in just the forest sciences, ecologists frequently differ on the most important resilience variable(s) (Nikinmaa et al. 2020) such that there is no obvious, single, “correct” way to choose resilience variable(s). A fourth and most important observation, however, is that there is no reason to expect that two variables in a given system will always (or even often) display qualitatively similar degrees of resilience to a given disturbance. Two resilience variables may be uncoupled so that one can display much more resilience than the other or even coupled negatively such that lower resilience in one variable facilitates the resilience of another (e.g., the example above in which a shift in phenology facilitates resilience of abundance). At the individual level, the concept of allostasis recognizes the fact that shifts in one variable are typically required to enable resilience in another variable (McEwen & Wingfield 2003). At larger biological scales, resilience variables within the same ecosystem often display very different degrees of resilience to the same perturbation. For example, specific functions, such as phenol oxidase C-cycling, very often responded in a qualitatively different manner than diversity metrics when soil microbe communities were experimentally perturbed (Orr et al. 2024). Together, the four observations above emphasize that conclusions about ecological resilience will be conditional on an investigator’s choice of which variables to measure . This key point implies a critical, and we believe underdiscussed, corollary statement, which is that it is likely impossible for any ecological system to be generally resilient, even to a single disturbance (press or pulse). Such “general resilience” would require a system to remain at and/or return to its pre-disturbance state in every meaningful variable. We speculate that such resilience may never occur in real ecological systems, even if one defines the pre-disturbance state of each variable broadly as some reasonable range of values, cyclic pattern of values, etc. Further, we note that general resilience is not demonstrated by any of our four case studies below. In the specific case of a press disturbance (as mentioned above), it is mathematically impossible for an ecological system to exhibit general resilience (assuming as always that the system is not 100% resistant; see Appendix for a general, mathematical proof); we refer to this principle as “Necessary Non-resilience”. If some variables recover to their pre-disturbance state despite the press disturbance, some other variables must necessarily be permanently shifted (i.e., not be fully resilient) to allow this recovery. This principle holds regardless of how complex the states of these variables may be (e.g., seasonal cycles) or the nature of the causal relationships among them (e.g., lags). One may raise the objection that the shifted variables may sometimes be considered ecologically unimportant while the resilient variables are ecologically important (e.g., if one considers plant abundance more important than plant phenology). To this objection we note that, if the shifted variables facilitate resilience in ecologically important variables, then the shifted variables should likely also be considered as ecologically important. To our knowledge, this Necessary Non-Resilience in at least one variable has not been previously pointed out in the ecological literature on resilience. Importantly, Necessary Non-Resilience explicitly challenges the common, implicit assumption in the ecological literature that one can define and assay overall ecological resilience; at least one important variable necessarily must be non-resilient (in the case of a press). Together, Necessary Non-Resilience to press disturbances and the less certain but likely lack of general resilience to pulse disturbances highlight the importance of variable diversity and variable choice. Rather than solely complicating the assessment of resilience for empirical systems, we instead argue in the sections below that explicit consideration of the conditionality of resilience, Necessary Non-resilience, and the relationships among resilience variables open potentially fruitful avenues for progress in resilience ecology. In particular, such consideration points a path towards the formulation of four research questions with explicit, testable hypotheses about which variables are more or less resilient within and across systems. The formulation of our research questions and hypotheses are aided by our conceptual model of the relationships among resilience variables across scales. Together, our observations suggest that a general understanding of ecological resilience requires understanding how different variables relate to each other in a system experiencing disturbance and, possibly, demonstrating subsequent resilience. This explicit focus on multivariate measurement of resilience, not just in terms of measuring multiple variables but in explicitly considering their differing responses and potential interactions, remains rare. Testable Resilience Hypotheses Given the expectation of differing resilience among variables, the categorization of variables, and a conceptual model linking them, we have motivation and means to ask which variables will be more resilient than others and when. We consider four broad research questions (RQ 1- 4), each of which engenders specific hypotheses. These hypotheses are scientifically interesting, testable, and potentially able to drive forward a general understanding of resilience across systems. Notably, we propose these hypotheses not because we believe them to necessarily be true, but because answering them would help to drive the science of resilience forward. We focus on questions relating to how factors, such as variable category, variable scale, or scale and connectivity of the ecological system, make variables more or less resilient. Notably, our research questions assume that variables within a category or scale are more similar in their resilience, which is also a testable hypothesis that our resilience variable categories are ecologically meaningful. We propose researching such relative resilience of different variables as a means to address the difficult diversity of resilience variables in three ways. First, we can quickly determine the importance of actively monitoring and managing the resilience of a given variable based on how likely it is to have lower or higher resilience, allowing us to focus our limited time on critical variables. Second, we would be able, with some confidence, to place lower or upper bounds on the resilience of variables we did not measure; e.g., if variables at smaller biological scales recover faster than those at larger biological scales, then measuring the time to recovery at the larger biological scale provides a probable upper bound on the time to recovery at the smaller biological scale. Third, deviations from the typical hierarchy of more or less resilient variables may highlight strong, uncharacterized interactions among variables worthy of novel investigations [similar to 24’s framework]. Thus, we value research questions and hypotheses regarding the relative resilience of different variables. RQ 1: What types of variables, within a scale, recover most quickly? An example hypothesis would be that, after a pulse disturbance, variables involving amounts recover quickly while variables involving characteristics or functions recover slowly (assuming both are perturbed). Specifically, we intend this hypothesis for population- and community-scale variables. For example, genetic diversity (a metric of a population-level characteristic) is often expected to recover more slowly after a bottleneck than population abundance; exactly this pattern was observed after a severe winter storm wiped out 95% of an isolated songbird population (Keller et al. 2001). To rigorously test this hypothesis, we would need sufficiently replicated measurements of recovering amounts and recovering characteristics from the same ecological systems at a given scale. Our rationale for why this hypothesis may be true, and therefore is worth testing, is that conspecific negative density dependence provides a broadly applicable, strong, stabilizing force (Johnson et al. 2012) that can drive recovery of amounts of a population. Scaling up, we expect this same process to drive recovery of amounts at the community level (Fig. 1). Characteristics and functions do not have as ubiquitous and rapid of a stabilizing force (e.g., evolution will rarely act as quickly as density dependences; see Appendix for further rationale related to full distributions vs. metrics). Whether or not our hypothesis is correct, we believe it is worth testing as a general understanding of whether and when amounts recover before characteristics would improve our understanding of the multivariate recovery of ecological systems. This hypothesis could be tested through monitoring the recovery time series of other variables, in addition to abundance, following a pulse disturbance. RQ2: Is there an order in which variables recover across scales? As a specific hypothesis, we propose that variables at smaller biological scales will display recovery before larger scales (assuming that both do recover). For example, this hypothesis predicts that individuals would most often recover biomass before population biomass recovers; population biomass recovery, in turn, would typically precede community level biomass recovery. Of course, we do not expect this hypothesis to apply if some scales are not recovering, e.g., an individual does not regain biomass but the population does due to immigration. When all scales under consideration do recover, we predict the recovery of smaller scales to precede larger scales for two reasons. First, smaller scales frequently turn over on faster time scales. Second, recovery at a smaller scale will often contribute to recovery at a larger scale if only by summation but recovery at a larger scale need not contribute to recovery at a smaller scale (see Fig. 1). Similar to knowing the order of recovery of variables within a scale, knowing the order of recovery of a single variable across scales will help us understand where in its multivariate recovery trajectory a system is and predict its future resilience. To further test this hypothesis, ecologists could explicitly track recovery trajectories across scales. These data exist already for certain variables [e.g., amounts of coral at different scales 24] and could be readily analyzed to test this hypothesis, e.g., via a meta-analysis across systems. RQ 3: How does biological scale affect the probability that a disturbed variable remains shifted instead of recovering? For example, we may hypothesize that disturbed characteristics are more likely to recover at the individual scale, less likely to recover at the population scale, and even less likely still to recover at the community scale. To test this hypothesis, investigators would need to measure the same disturbed variable across scales, say metabolic rate, across time and determine the proportion of recoveries, across replicate units, for the variable at the individual scale (e.g., an individual’s specific leaf area: von Arx et al. 2012), population scale (population average specific leaf area), and at the community scale (community average specific leaf area: McCoy-Sulentic et al. 2017). Observation of recovery vs. a shift would need to be defined within a relevant time period defined by the investigator. If the proportion of recoveries was higher at the individual scale than population scale and higher at the population scale than community scale, the data would support this hypothesis. We believe this hypothesis may be true due to the strength of homeostasis, driving characteristics to recover at the individual scale. While the power of homeostasis to drive recovery likely aggregates to higher scales, it will be diluted by other processes such that the larger biological scales are less likely to experience recovery. For example, in our case studies below, host individuals commonly recover from infection (thus recovery in the infection status characteristic) while host populations rarely recover to the low or zero percent infection status that it had before the disturbance of disease invasion. Variables represented as amounts, however, may show different patterns than those measured as characteristics. Amounts, we hypothesize, may be more likely to shift (i.e., less likely to recover) at lower scales than higher scales (the inverse of our hypothesis for characteristics). We propose this because the most powerful driver of a permanent shift in amounts is death (or extinction at higher scales) and larger groups are statistically less vulnerable to extinction (Iwasa et al. 2000). In our case studies, we see loss of individuals, occasional loss of populations, but no instances of loss of entire communities. Understanding the relative resilience of variables of different types across biological scales will prove informative for future management decisions as measuring the resilience of a variable at one biological scale may help infer its resilience at other scales. RQ 4: How does dispersal affect resilience across scales and variables? Unlike the previous questions that emphasize the effect of variable scale or category, now we consider the effect of an ecosystem property on resilience. The metapopulation literature has extensively considered how dispersal, the rate of movement of individuals among different patches, affects metapopulation persistence (a way of quantifying resilience in amounts, typically to stochastic pulse disturbances; Ovaskainen & Saastamoinen 2018). Dispersal can increase metapopulation persistence in a spatiotemporally heterogeneous environment but too much dispersal may decrease persistence by synchronizing population fluctuations (Earn et al. 2000) or spreading infectious diseases (Hess 1996) so that persistence is often maximized at intermediate dispersal (Ben Zion et al. 2010). We hypothesize that intermediate dispersal promotes resilience. Such findings have also been found at the metacommunity level in that intermediate dispersal often maximizes species diversity, though not always (Haegeman & Loreau 2014), representing trends in the resilience of the species diversity variable to the stochastic pulse disturbances. Less attention has explicitly connected such a hypothesis to the individual scale, but we believe there is reason to expect this hypothesis may hold true for individuals. Movement across a spatiotemporally heterogeneous landscape allows an individual to buffer against stochastic pulse disturbances, e.g., by behavioral thermoregulation (Goller et al. 2014) so that more dispersal at the individual level improves the resilience of individual body temperature to pulse disturbances. But too much movement through heterogeneous environments is both energetically costly and may reduce acclimation to local conditions, leading to stress (e.g. to salinity; Dildar et al. 2025). Since fitness should decline at the extremes of dispersal, we believe it is worth testing whether intermediate dispersal promotes resilience at individual, population, and community scales, perhaps through cross-scale interactions, or whether dispersal-resilience patterns differ strongly across scales and/or resilience variables. Case Studies Our categorization and conceptual model of diverse resilience variables stimulates new research questions and hypotheses. We argue that these hypotheses are useful because it is important to know if they are true and because they are testable. Below, we argue for the empirical relevance of these hypotheses by outlining four empirical case studies that each relate, with varying strengths, to at least two of our Necessary Non-resilience argument and the hypotheses corresponding to our research questions (RQ1-4). Mountain yellow-legged frogs and chytrid fungus The fungal pathogen Batrachochytrium dendrobatidis ( Bd ) arrived in the Sierra Nevada mountains of California, USA in the 1970s and has since caused devastating declines in Mountain yellow-legged frog (MYL frog) populations. In combination with other stressors such as introduced trout and pesticides, Bd has contributed to reducing the historic MYL frog range by 95% (Knapp et al. 2016). While hundreds of populations of MYL frogs have been extirpated from the Sierra Nevada’s due to Bd -induced declines, some populations have persisted despite suffering Bd -induced declines and abundances are beginning to recover (Knapp et al. 2016). Here, we consider two observations regarding resilience in the MYL frog-Bd system and how they relate to our framework and hypotheses above (Box 1). | Box 1. MYL frog key observations. Necessary Non-resilience: A permanent shift in one variable facilitates recovery in another: To date, the key variable of interest in assessing the resilience of MYL frogs to Bd disturbance has been frog abundance (particularly, the number of adult frogs in a population). Frog abundance is recovering in populations – we observed drastic declines (i.e., shifts) in abundance following Bd arrival (Vredenburg et al. 2010) followed by ongoing recovery in some sites (Joseph & Knapp 2018; Knapp et al. 2016, 2024). There is also emerging evidence that resilience in this metric came as a result of a trade-off – epidemiological and genetic data suggest that recovering hosts have evolved resistance to Bd (Byrne et al. 2025). Epidemiologically, recovering hosts are still infected at high prevalence, but the intensity of Bd on frogs is reduced compared to naive frogs that succumb to chytridiomycosis (Knapp et al. 2016, 2024), consistent with intensity-reduction resistance mechanisms. Genetically, compared to naive frog populations, there are consistent signatures of selection on regions of the genome related to skin defenses in populations of frogs that are recovering with Bd (Byrne et al. 2025). These results highlight an example of where resilience along one population-level metric (abundance is recovering) is likely a result of lack of resilience along another population-level metric (frequency of particular skin defense alleles in the population have permanently changed). In this case, the “lack of resilience” in allele frequency is in many ways beneficial as it facilitates resilience in a metric that managers and the public care more about (e.g., frog abundance). RQ 2: Smaller scales recover faster: In addition to the evolution of intensity-reduction resistance from standing variation, another hypothesized mechanism of recovery in the MYL frog system is due to acquired immunity following repeated Bd infections (Briggs and Toothman, unpublished data). In the lab, when frogs are able to clear initial infections, they have a lower probability of being infected again given exposure, have lower infection loads given infection, and higher survival given infection. Thus, the mortality hazard from Bd infection (an individual-level characteristic, see Table 1), determined from proxies such as infection load, goes up with initial Bd invasion then reduces with acquired immunity for a given frog. The time scale of this acquired immunity (i.e., individual resilience) is rapid, with frogs developing and maintaining immunity within 1 - 6 months of the initial exposure. In contrast, the vital rates of recovering MYL frog populations, a population-level characteristic directly connected to individual-level hazard) likely recovers more slowly over the scale of decades (abundance recovers on the scale of decades: Knapp et al. 2016). The quantitative importance of acquired immunity to population recovery is an ongoing question in this system but capture-mark-recapture data may be able to confirm that individual-level recovery is preceding population-level recovery. | White-Nose Syndrome in bats Our second case study is focused on another fungal pathogen Pseudogymnoascus destructans (Pd )—the causative agent of White-nose Syndrome (WNS)—which has led to severe declines in bat populations along the Eastern and Central United States. The little brown bat (Myotis lucifugus, LBB) is one of the most severely impacted species in North America, experiencing up to 90-100% declines in infected populations (Cheng et al. 2021). However, other species have been more tolerant or resistant to the effects and have experienced more moderate population-level effects, such as the big brown bat (Eptesicus fuscus, BBB). Although BBB can still become infected and die from Pd, their populations generally are not as severely impacted as other species (Cheng et al. 2021; Lemieux-Labonté et al. 2020). In fact, even in the presence of Pd, evidence suggests some BBB populations could increase. In Southern Ontario, acoustic monitoring data indicated that BBB populations may actually be increasing (Morningstar et al., 2019). Researchers suspect this may be due, in part, to competitive release by the decrease in other bat species populations, namely the LBB. LBB and BBB share common resources such as food and hibernacula. As LBB numbers dwindle, there are more resources for BBB. As with any significant change in species population sizes, there are inevitably broader ecological implications. For instance, declining insectivorous bat populations (e.g., LBB) may impact the insect populations in an area, with the expectation that insect populations would increase due to the reduction of natural predators. Morningstar et al. (2019) instead demonstrated a potential means of resilience in this metric with a shift in heterospecific behavior due to the competitive release between LBB and BBB. Not only does the evidence suggest that BBB populations are increasing, possibly acting to compensate for the decrease in LBB by eating more insects, but examination of BBB guano indicates that this species has expanded its diet to incorporate insect species that were primarily consumed only by LBB. Notably, this study (Morningstar et al. 2019) does not report on changes to insect populations—the proposed resilience metric. Rather, it provides evidence for shifted individual and species-level characteristics that may facilitate resilience on a community level. This system supports several of our key points relating to resilience (Box 2). | Box 2. Little and Big Brown Bat key observations. Necessary Non-resilience: A permanent shift in one variable facilitates recovery in another. In this case study, the size and even distribution of the insect community are resilient metrics. Given a disturbance—the introduction of Pd —likely all bat populations experienced some degree of decline. With fewer predators, insect populations could flourish and increase in abundance (temporary shift). However, since BBBs were not as negatively impacted by the disease, they had the opportunity to fill the niche left behind by the dwindling LBB populations. Recovering their numbers due to increased resources means that the BBBs also demonstrated resilience, and to a bat biologist, this may be the conclusion. However, in the context of the community’s recovery, BBB population recovery to greater than pre-disturbance states would be a shift that enables the species to consume more insects than before Pd and reduce insect community size back to its pre-disturbance state. In this case, the insect community abundance is resilient, declining back to its pre-disturbance state. Shifts in bat population sizes—increase in BBB after LBB decreased—enabled resilience in insect community abundance. Another shift observed was in BBB eating habits. They expanded their diets to incorporate insect species that had, prior to the disturbance, been consumed by LBB. Not only do increased BBB populations control insect community size, but their behavioral shifts likely impact insect species community structure (species abundance distribution). RQ 2: Smaller scales recover faster. In this case study, one bat population (BBB) recovers its numbers or at least population-level biomass after an initial depression. BBB are less susceptible to the ill-effects of WNS, and they have access to more resources as other bat populations in the community are reduced. Therefore, even as some population biomasses are continuing to decrease, BBB biomass is increasing and recovering. At the same time, on the community scale, bat biomass decreases with the initial epizootic. The increasing BBB biomass increases or at least slows the loss of biomass on the community level. If overall community biomass does indeed recover, it will occur more slowly than the population recovery (for those populations that do, in fact, recover their biomass). | Coral reefs of Moorea Our third case study is the coral reef community on the island of Moorea, French Polynesia. Unlike the previous two systems that involved resilience to the press disturbance of invasion of infectious diseases, this example illustrates the response of the coral reef system to repeated pulse disturbances from different sources. Coral reefs involve strong competitive interactions between corals and macroalgae. The primary state variable of interest is frequently the percent of space on the hard substrate that is covered by corals versus macroalgae versus other benthic space holders (i.e., coral cover). State shifts in other coral reef systems (e.g. much of the Caribbean following the die-off of herbivorous Diadema sea urchins in the 1980s) have led to wide-spread shifts from coral-dominated to algal-dominated states (Lessios 2016; Levitan et al. 2023). Therefore, there is strong interest in the mechanisms by which some reefs are able to bounce back to a coral-dominated state following disturbances (Edmunds et al. 2018; Gouezo et al. 2019). Additional state variables of interest include the abundance, biomass, and species composition of the fish communities that can control the macroalgal and coral communities, and the species composition of the coral and algae communities. In 2007-2009, an outbreak of crown of thorns seastars (COTS), which eat corals, led to wide-spread loss of live corals from Moorea. COTS kill live corals but leave the coral skeletons largely intact. This is important for the resilience of the coral community, because both corals and macroalgae recruit to the structures of the coral skeletons, however, in the absence of sufficient grazing by herbivorous fishes, this structure can quickly become dominated by macroalgae (Holbrook et al. 2016; Kopecky et al. 2025). In 2010 the island of Moorea was hit by cyclone Oli, which broke up and removed much of the remaining coral structure. More recently, in 2019, the region experienced periods with extremely high temperatures, and the coral bleaching occurred on Moorea. In a bleaching event, the symbiotic dinoflagellates in the corals are killed, resulting in death of the coral tissue, but leaving the coral skeletons intact (Kopecky et al. 2023; Winslow et al. 2024). The resilience, or lack thereof, of these ecosystems connect to some of our key points (Box 3). | Box 3. Coral key observations. RQ 1: After a pulse, amounts recover faster. Due to demographic differences, populations of some species of corals were able to rebound from disturbances faster than others, and it can take a substantial amount of time to recover the pre-disturbance coral community structure (Gouezo et al. 2019). In particular, branching corals, such as Pocillopora, have life-history traits, including high reproductive rates, rapid recruitment and fast growth that allow them to increase rapidly in abundance (and percent cover). Their branching morphology, however, makes these corals more vulnerable to mechanical disturbances (such as cyclones). In contrast, mounding corals, such as Porites) grow more slowly, but are much more resistant to physical damage due to their compact, dome-shaped morphology (Edmunds et al. 2018; Holbrook et al. 2018). Thus, many community-scale amounts such as total coral biomass, total coral cover, etc. recovered more quickly than community-scale characteristics such as mean structural complexity. RQ 4: Intermediate dispersal promotes resilience. Following the COTS/cyclone combined disturbance event, two types of spatial variation were observed in the rate of return to the coral-dominated state. First, the rate of recovery varied greatly among different sides of Moorea that are exposed to different oceanographic conditions (Holbrook et al. 2018). Corals exhibit open recruitment, in which the coral propagules that settle on a given reef can originate from corals throughout the region, rather than being produced locally. In contrast, most macroalgal reproduction occurs locally (although long-distance dispersal can occur from algal fragments that break off and drift to other reefs: Adam et al. 2022; Adjeroud 1997). Spatial heterogeneity in the recruitment of coral propagules drove the spatial variation in the rate of recovery of coral cover in different sides of the island Moorea, with areas that received the highest quantities of dispersing corals (due to ocean circulation patterns and other oceanographic factors) recovering to the coral-dominated state fastest (Holbrook et al. 2018). | Case study 4: Resilience of the American pika to climate change The American pika, Ochotona princeps, a small mammal typically restricted to mid-high elevation alpine regions, is an excellent case study to understand potential resilience to climate change. Pikas are adapted to cold climates, and some populations in northern California and the Great Basin in southern Utah have been extirpated [49]. These sites are warmer and drier than optimal for this species and have been impacted by cattle grazing and other anthropogenic impacts. Despite being largely restricted to moist, rocky, mountainous locations, occupying large mountain ranges in North America, recent work indicates that pikas demonstrate behavioral (Beever et al. 2017) and physiological (Solari & Hadly 2020) plasticity in response to changing environments, which can buffer them from the effects of climate change. In addition, pikas demonstrate signatures of climate-mediated genetic changes which may indicate adaptation to new conditions (Sjodin et al. 2024; Sjodin & Russello 2022). Altogether, the resilience of pikas connects to two of our key points (Box 4). | Box 4. Pika key observations.Necessary non-resilience: A permanent shift in one variable facilitates recovery in another. Pikas are sensitive to high temperatures, maintain body temperatures (40.1˚C) close to their critical thermal maximum (43.1˚C), and have a limited ability to physiologically dissipate heat (Smith & Weston 1990). To survive in lower elevation, marginal habitat, pikas demonstrate considerable behavioral flexibility, retreating to inside rocky talus habitat during the heat of the day, and increasing nighttime foraging activity (Beever et al. 2017). In addition, while pika have been extirpated at some fragmented, low-quality sites, they have also been found persisting in and sometimes recolonizing atypical low elevation habitats, such as anthropogenic ore dumps near volcanic rock outcroppings near Bodie, California, and temperate rainforest in the Columbia River Gorge in Oregon, which is near sea-level (Millar & Smith 2022; Varner & Dearing 2014).Furthermore, pikas rely on large caches of food to sustain them through the winter, which they build up through typical haying behavior. In these marginal sites where vegetation is sparser, pikas have been shown to spend more time foraging and less time haying, as winter conditions are less harsh (Smith 2020). In addition, pikas in non-typical habitats have also demonstrated flexibility in their diet, consuming a wide variety of vegetation types not usually found in alpine locations (Smith 2020). For example, in the Columbia River Gorge populations, pika diet is up to 60% moss (Varner et al. 2016; Varner & Dearing 2014). This shift in thermoregulatory behavior and diet at the population scale lends support for our argument of Necessary Non-resilience, in which pika populations have shifted their behavior to recover in abundance in typically non-optimal sites in the face of climate change-driven aridification and warming.In addition, recent studies provide evidence that pika populations can adapt to new climatic conditions. A range-wide study found hundreds of positively selected genes associated with hypoxia, cold, and UV tolerance, which would benefit pika moving up mountains to track their fundamental environmental niche (Sjodin & Russello 2022). There is also evidence of increased copy number variation of genes correlated with gradients in temperature, precipitation, and solar radiation, which is indicative of local adaptation to climate (Sjodin et al. 2024). Finally, studies in Himalayan pikas have found considerable physiological flexibility in hypoxia tolerance, allowing movement to higher elevations, which warrants further study in the American pika (Solari & Hadly 2020). Putative adaptation to new climatic conditions at the population scale lends support for necessary non-resilience as well.RQ 4: Intermediate dispersal promotes resilienceThis flexibility in their behavior and diet, as well as evidence of local adaptation, provides evidence that pika may be able to withstand changing climatic conditions better than previously thought. This plasticity shifts pika ecology and behavior, while selection leads to better adapted individuals surviving new climates, ultimately resulting in resilience at the population scale. However, preventing pika extirpation in marginal habitat, particularly as global average air temperatures continue to increase, will require connectivity with upslope habitat, as these small mammals are poor dispersers (Franken & Hik 2004; Smith 2020) and are demonstrating evidence of genetic isolation in the eastern Sierra Nevada and Great Basin populations in Nevada (Klingler et al. 2021). The importance of dispersal to Pika population persistence aligns with our hypothesis that populations linked to larger networks are more likely to recover after disturbance than isolated populations. Overall, pikas are demonstrating resilience to the on-going press disturbance of climate change through potential plasticity and/or local adaptation in behavior and physiology at the population scale. | Overall, our case studies demonstrate three important facts. First, different resilience variables often respond very differently to the same disturbance. Second, these differences are often related to the fact that a shift in one variable is critical to the recovery of another variable, similar to our Necessary Non-resilience argument. Third, our case studies demonstrate that our research questions and hypotheses are biologically relevant and empirically testable, though doing so rigorously across different variable categories and scales would require significantly more data and likely meta-analyses. These facts emphasize that ecologists must embrace the diversity of resilience variables and that doing so can highlight important new research directions.

Ecological resilience variables can be incredibly diverse in their definitions and responses to the same disturbance. We categorized this diversity into Amounts, Characteristics, and Functions at the individual, population, and community scales (though other scales can easily be used). We argued that variables will often qualitatively differ in whether or not they were resilient to a disturbance, emphasized by our Necessary Non-resilience argument showing that at least one variable will necessarily be non-resilient to a press disturbance. These differing responses can be critical as ecologists, stakeholders, and policymakers differ significantly, within and among fields, regarding which resilience variables and at which biological scales are most important. In the absence of a clear, objective way to determine variable importance, we advocate for explicitly acknowledging the diversity of resilience variables while attempting to characterize their interactions and differential resilience to disturbance. Additionally, we provided a conceptual model for the relationships among resilience variable categories that facilitates the formation of research questions and hypotheses. We focused on four, testable questions and hypotheses that arise from our framework and discussed how they, and the Necessary Non-resilience argument, have varying degrees of support from four case studies. In summary, we argue for explicitly acknowledging variable diversity and grappling with it through categorization and explicit conceptual models, and the formation of explicit questions and hypotheses about these variable categories that can be tested in real, empirical systems (such as those of our case studies). Ecological theory is both partially to blame for our sparse-variable thinking and also a critical part of the path forward for future work. Mathematical theory has developed very useful tools for predicting, measuring, and understanding the resilience of a single variable at a time such as basin width, mean exit time, critical slowing down, etc. (Dakos & Kéfi 2022). Such theory may have encouraged ecologists to consider the resilience of a single variable. However, the theoretical tools exist to consider multiple variables or multiple axes representing many collapsed variables and other questions of diverse variables (Nolting & Abbott 2016). Further, existing theory has encouraged considering the interactions of variables, particularly across scales (Gunderson & Holling 2002). Traditionally, the existing multidimensional tools have been used to study the overall resilience of a basin (Dakos & Kéfi 2022; Nolting & Abbott 2016), implicitly focusing on whether a system is resilient in all of the important variables. New routes forward must consider the importance of different resilience responses of different variables and their interaction, particularly for understanding resilience to the next disturbance. For example, if an ecological system moves into another basin of attraction through a change in one variable, are there any empirically relevant, theoretical rules for how that affects the resilience of other variables (e.g., width of the new basin in the dimensions that were resilient)? This question relates to the differences between local and global stability analysis and is also closely related to the critical frontier of cascading regime shifts, in which one regime shift can increase the probability of another (Rocha et al. 2018). Further, do these rules follow patterns for certain types of variables, e.g., does non-resilience in amounts have consistent effects on the resilience of characteristics? Future theory-data integration can help us build toward a more robust, general understanding of ecological systems with combinations of variables that display widely different degrees of resilience. Such understanding will critically guide future management as we acknowledge ecological systems will not be generally resilient to the myriad disturbances of global change and, therefore, we must predict how the failure of resilience in one variable affects the resilience of other variables. In addition to the importance of variable diversity, we note the importance of the ongoing consideration of disturbance diversity. Previous authors have recognized different disturbance types beyond pulse/press such as flow-kick (Reed et al. 2024) or noise (Kéfi et al. 2019). But biotic agents such as invasive species and emerging infectious diseases (which may constitute or synergize with invasive species: Strauss et al. 2012) pose increasingly ubiquitous threats to ecological systems (Diagne et al. 2021). Frequently, the biotic agent is considered as the disturbance more than as a member of the disturbed community (Guo et al. 2018). While many of the emerging labels for different types of disturbances refer to how their intensity may change with time, biotic disturbances often possess the unique property of responding to the disturbed ecosystem. Biotic disturbances can respond to the disturbed ecosystem through evolution, plasticity, or population dynamics. For example, disturbed host populations decline in abundance, in turn reducing the abundance of infectious pathogens, a dynamic that fundamentally structures how pathogens can and cannot drive host populations extinct (De Castro & Bolker 2005); we typically think of larger populations as generally more resilient in terms of lower extinction risk to various disturbances (Purvis et al. 2000) but they may also be more susceptible to the rapid spread of a novel infectious disease (Dallas et al. 2018). Beyond parasitic interactions, competitors may serve as biotic disturbances, e.g., native plants may evolve resistance to invaders’ allelochemicals, competing more effectively (Strauss et al. 2006) and potentially reducing invader abundance. These strong feedbacks may affect the principles of ecological resilience in critical, underappreciated ways. Thus, we add to the ongoing conversation regarding the diversity of disturbance types by emphasizing the need to know when biotic disturbances do and do not follow the same rules of resilience ecology as abiotic disturbances. Both diverse disturbance types and diverse resilience variables make it more difficult to advance resilience ecology in a general manner, but we offer some recommendations. First, investigators can be more transparent about variables they measured and did not measure while avoiding sweeping statements regarding overall resilience. For example, the rate of recovery in native plant cover (Hensel et al. 2021), native plant biomass (van Der Loop et al. 2023), ecosystem services (Nyssen et al. 2024), have each been implied to be the obvious, relevant resilience variable in an ecosystem resilience framework. We suggest that it would be beneficial to more explicitly describe resilience as being in the measured variable while discussing possible implications for other variables as these variables may respond differently. Second, investigators who measure multiple variables can report how those variables displayed similar or very different degrees of resilience. Often, these diverse variables are collapsed into metrics that are used to assay overall resilience or lack thereof (Lamothe et al. 2019; Li et al. 2023; Li & Wang 2023). We call for more investigators to consider the differential resilience responses of uncollapsed but categorized variables, the implications of these differential responses, and the additional information that can be obtained from them [e.g., via the methods in 9]. Third, investigators can be explicit about their conceptual (or mathematical) model of the relationships among resilience variables when arguing for the relevance of their chosen resilience variables. These recommendations apply to chosen combinations of collapsed variables just like individual variables, arguing for the relevance of a single, collapsed metric of ecological resilience (e.g., as used by Li et al. 2023; Li & Wang 2023) while acknowledging the information missed in such an approach. Explicitly grappling with variable diversity will provide new routes forward in resilience ecology where general progress has been hindered by investigators choosing different variables while speaking of “resilience” broadly. Ecologists must recognize the importance of diverse and differentially responding variables to stimulate productive new questions and hypotheses. What kinds of variables and systems are more resilient than others? If we can answer that question, we will understand when measuring a subset of ecological variables informs resilience about other, unmeasured variables. We offer a possible categorization of resilience variables that facilitates four, practical research questions with testable hypotheses for which kinds of variables and systems are more resilient that others. Our four case studies support the immediate, empirical relevance and testability of our hypotheses. Testing these hypotheses, and others yet to be formulated, about the relationships among diverse resilience variables will provide a more robust, general understanding of resilience ecology. We argue that explicitly grappling with resilience variable diversity, such as through methods discussed above, is more likely to aid resilience ecology than attempting to build a general understanding based on investigating a small number of variables at a time. And we need a more robust, general understanding of resilience ecology to guide scientists, stakeholders, and policymakers as diverse variables and ecological systems face growing numbers of disturbances. Some important variables will inevitably fail to be resilient but with the proper understanding of resilience variables, we can more adequately predict these and gauge their ramifications for other aspects of ecological resilience.

We thank Roland Knapp, Thomas Smith, Louise Rollins-Smith, Ruijiao Sun, Jake Eisaguirre, Brandon Hoenig, and Gui Becker for their formative discussions, particularly early in the development of this manuscript.

Adam, T.C., Holbrook, S.J., Burkepile, D.E., Speare, K.E., Brooks, A.J., Ladd, M.C., et al. (2022). Priority effects in coral–macroalgae interactions can drive alternate community paths in the absence of top‐down control. Ecology, 103, e3831. Adjeroud, M. (1997). Factors influencing spatial patterns on coral reefs around Moorea, French Polynesia. Mar. Ecol. Prog. Ser., 159, 105–119. von Arx, G., Archer, S.R. & Hughes, M.K. (2012). Long-term functional plasticity in plant hydraulic architecture in response to supplemental moisture. Ann. Bot., 109, 1091–1100. Beever, E.A., Hall, L.E., Varner, J., Loosen, A.E., Dunham, J.B., Gahl, M.K., et al. (2017). Behavioral flexibility as a mechanism for coping with climate change. Front. Ecol. Environ., 15, 299–308. Ben Zion, Y., Yaari, G. & Shnerb, N.M. (2010). Optimizing metapopulation sustainability through a checkerboard strategy. PLoS Comput. Biol., 6, e1000643. Byrne, A., Rothstein, A., Knapp, R. & Rosenblum, E. (2025). Genomic correlates of disease recovery in natural populations of mountain yellow-legged frogs. Authorea . Capdevila, P., Stott, I., Oliveras Menor, I., Stouffer, D.B., Raimundo, R.L., White, H., et al. (2021). Reconciling resilience across ecological systems, species and subdisciplines. J. Ecol., 109, 3102–3113. Cheng, T.L., Reichard, J.D., Coleman, J.T., Weller, T.J., Thogmartin, W.E., Reichert, B.E., et al. (2021). The scope and severity of white‐nose syndrome on hibernating bats in North America. Conserv. Biol., 35, 1586–1597. Cleland, E.E., Chuine, I., Menzel, A., Mooney, H.A. & Schwartz, M.D. (2007). Shifting plant phenology in response to global change. Trends Ecol. Evol., 22, 357–365. Dakos, V. & Kéfi, S. (2022). Ecological resilience: what to measure and how. Environ. Res. Lett., 17, 043003. Dallas, T.A., Krkošek, M. & Drake, J.M. (2018). Experimental evidence of a pathogen invasion threshold. R. Soc. Open Sci., 5, 171975. Davies, T.W., Jenkins, S.R., Kingham, R., Hawkins, S.J. & Hiddink, J.G. (2012). Extirpation‐resistant species do not always compensate for the decline in ecosystem processes associated with biodiversity loss. J. Ecol., 100, 1475–1481. De Castro, F. & Bolker, B. (2005). Mechanisms of disease‐induced extinction. Ecol. Lett., 8, 117–126. van Der Loop, J.M., van Kleef, H.H., van Veenhuisen, L.S., Lamers, L.L. & Leuven, R.S. (2023). The ecosystem resilience approach to control the invasive alien species Australian swamp stonecrop (Crassula helmsii). Restor. Ecol., 31, e13844. Diagne, C., Leroy, B., Vaissière, A.-C., Gozlan, R.E., Roiz, D., Jarić, I., et al. (2021). High and rising economic costs of biological invasions worldwide. Nature, 592, 571–576. Dildar, T., Cui, W., Ikhwanuddin, M. & Ma, H. (2025). Aquatic Organisms in Response to Salinity Stress: Ecological Impacts, Adaptive Mechanisms, and Resilience Strategies. Biology, 14, 667. Earn, D.J., Levin, S.A. & Rohani, P. (2000). Coherence and conservation. Science, 290, 1360–1364. Franken, R.J. & Hik, D.S. (2004). Influence of habitat quality, patch size and connectivity on colonization and extinction dynamics of collared pikas Ochotona collaris. J. Anim. Ecol., 889–896. Goller, M., Goller, F. & French, S.S. (2014). A heterogeneous thermal environment enables remarkable behavioral thermoregulation in Uta stansburiana. Ecol. Evol., 4, 3319–3329. Gunderson, L.H. & Holling, C.S. (2002). Panarchy: Understanding transformations in human and natural systems. Guo, Q., Brockway, D.G., Larson, D.L., Wang, D. & Ren, H. (2018). Improving ecological restoration to curb biotic invasion—A practical guide. Invasive Plant Sci. Manag., 11, 163–174. Haegeman, B. & Loreau, M. (2014). General relationships between consumer dispersal, resource dispersal and metacommunity diversity. Ecol. Lett., 17, 175–184. Hensel, M.J., Silliman, B.R., van de Koppel, J., Hensel, E., Sharp, S.J., Crotty, S.M., et al. (2021). A large invasive consumer reduces coastal ecosystem resilience by disabling positive species interactions. Nat. Commun., 12, 6290. Hess, G. (1996). Disease in metapopulation models: implications for conservation. Ecology, 77, 1617–1632. Hodgson, D., McDonald, J.L. & Hosken, D.J. (2015). What do you mean,‘resilient’? Trends Ecol. Evol., 30, 503–506. Hoelzel, A.R., Gkafas, G.A., Kang, H., Sarigol, F., Le Boeuf, B., Costa, D.P., et al. (2024). Genomics of post-bottleneck recovery in the northern elephant seal. Nat. Ecol. Evol., 8, 686–694. Holbrook, S.J., Adam, T.C., Edmunds, P.J., Schmitt, R.J., Carpenter, R.C., Brooks, A.J., et al. (2018). Recruitment drives spatial variation in recovery rates of resilient coral reefs. Sci. Rep., 8, 1–11. Holbrook, S.J., Schmitt, R.J., Adam, T.C. & Brooks, A.J. (2016). Coral reef resilience, tipping points and the strength of herbivory. Sci. Rep., 6, 35817. Holling, C.S. (1973). Resilience and stability of ecological systems . International Institute for Applied Systems Analysis Laxenburg. Iwasa, Y., Hakoyama, H., Nakamaru, M. & Nakanishi, J. (2000). Estimate of population extinction risk and its application to ecological risk management. Popul. Ecol., 42, 73–80. Johnson, D.J., Beaulieu, W.T., Bever, J.D. & Clay, K. (2012). Conspecific negative density dependence and forest diversity. Science, 336, 904–907. Joseph, M.B. & Knapp, R.A. (2018). Disease and climate effects on individuals drive post‐reintroduction population dynamics of an endangered amphibian. Ecosphere, 9, e02499. Kéfi, S., Domínguez‐García, V., Donohue, I., Fontaine, C., Thébault, E. & Dakos, V. (2019). Advancing our understanding of ecological stability. Ecol. Lett., 22, 1349–1356. Keller, L.F., Jeffery, K.J., Arcese, P., Beaumont, M.A., Hochachka, W.M., Smith, J.N., et al. (2001). Immigration and the ephemerality of a natural population bottleneck: evidence from molecular markers. Proc. R. Soc. Lond. B Biol. Sci., 268, 1387–1394. Klingler, K.B., Jahner, J.P., Parchman, T.L., Ray, C. & Peacock, M.M. (2021). Genomic variation in the American pika: signatures of geographic isolation and implications for conservation. BMC Ecol. Evol., 21, 2. Knapp, R.A., Fellers, G.M., Kleeman, P.M., Miller, D.A., Vredenburg, V.T., Rosenblum, E.B., et al. (2016). Large-scale recovery of an endangered amphibian despite ongoing exposure to multiple stressors. Proc. Natl. Acad. Sci., 113, 11889–11894. Knapp, R.A., Wilber, M.Q., Joseph, M.B., Smith, T.C. & Grasso, R.L. (2024). Reintroduction of resistant frogs facilitates landscape-scale recovery in the presence of a lethal fungal disease. Nat. Commun., 15, 9436. Kopecky, K.L., Pavoni, G., Corsini, M., Brooks, A.J., DiFiore, B.P., Menna, F., et al. (2025). Removing dead coral after marine heatwaves can mitigate coral–algae competition and increase viable coral recruitment. Ecol. Appl., 35, e70077. Kopecky, K.L., Pavoni, G., Nocerino, E., Brooks, A.J., Corsini, M., Menna, F., et al. (2023). Quantifying the loss of coral from a bleaching event using underwater photogrammetry and AI-assisted image segmentation. Remote Sens., 15, 4077. Krueger, K.A. (n.d.). ‘Unsuitable for release’ | TWRA responds after deciding to euthanize 13 bear cubs. WVLT8 . Lachs, L., Biondi, P., Gouezo, M., Nestor, V., Olsudong, D., Guest, J., et al. (2024). Demographic recovery of corals at a wave-exposed reef following catastrophic disturbance. Coral Reefs, 43, 193–199. Lamothe, K.A., Somers, K.M. & Jackson, D.A. (2019). Linking the ball‐and‐cup analogy and ordination trajectories to describe ecosystem stability, resistance, and resilience. Ecosphere, 10, e02629. Lemieux-Labonté, V., Dorville, N.A.-Y., Willis, C.K. & Lapointe, F.-J. (2020). Antifungal potential of the skin microbiota of hibernating big brown bats (Eptesicus fuscus) infected with the causal agent of white-nose syndrome. Front. Microbiol., 11, 1776. Lessios, H.A. (2016). The great Diadema antillarum die-off: 30 years later. Annu. Rev. Mar. Sci., 8, 267–283. Levitan, D.R., Best, R.M. & Edmunds, P.J. (2023). Sea urchin mass mortalities 40 y apart further threaten Caribbean coral reefs. Proc. Natl. Acad. Sci., 120, e2218901120. Li, D., Yang, W. & Huang, R. (2023). The multidimensional differences and driving forces of ecological environment resilience in China. Environ. Impact Assess. Rev., 98, 106954. Li, G. & Wang, L. (2023). Study of regional variations and convergence in ecological resilience of Chinese cities. Ecol. Indic., 154, 110667. Martins, I.S., Schrodt, F., Blowes, S.A., Bates, A.E., Bjorkman, A.D., Brambilla, V., et al. (2023). Widespread shifts in body size within populations and assemblages. Science, 381, 1067–1071. McCoy-Sulentic, M.E., Kolb, T.E., Merritt, D.M., Palmquist, E., Ralston, B.E., Sarr, D.A., et al. (2017). Changes in community-level riparian plant traits over inundation gradients, Colorado River, Grand Canyon. Wetlands, 37, 635–646. McEwen, B.S. & Wingfield, J.C. (2003). The concept of allostasis in biology and biomedicine. Horm. Behav., 43, 2–15. Medin, D.L. & Bang, M. (2014). Who’s asking?: Native science, western science, and science education . MIT Press. Millar, C.I. & Smith, A.T. (2022). Return of the pika: American pikas re-occupy long-extirpated, warm locations. Ecol. Evol. Mori, A. (2016). Resilience in the studies of biodiversity–ecosystem functioning. Trends Ecol. Evol., 31, 87–89. Newman, E.A., Wilber, M.Q., Kopper, K.E., Moritz, M.A., Falk, D.A., McKenzie, D., et al. (2020). Disturbance macroecology: a comparative study of community structure metrics in a high‐severity disturbance regime. Ecosphere, 11, e03022. Nikinmaa, L., Lindner, M., Cantarello, E., Jump, A.S., Seidl, R., Winkel, G., et al. (2020). Reviewing the use of resilience concepts in forest sciences. Curr. For. Rep., 6, 61–80. Nolting, B.C. & Abbott, K.C. (2016). Balls, cups, and quasi‐potentials: quantifying stability in stochastic systems. Ecology, 97, 850–864. Nyssen, B., Ouden, J.D., Bindewald, A., Brancalion, P., Kremer, K., Lapin, K., et al. (2024). Established invasive tree species offer opportunities for Forest Resilience to Climate Change. Curr. For. Rep., 10, 456–486. Odum, E.P. & Barrett, G.W. (1971). Fundamentals of ecology. Orr, J.A., Piggott, J.J., Jackson, A.L., Jackson, M.C. & Arnoldi, J.-F. (2024). Variability of functional and biodiversity responses to perturbations is predictable and informative. Nat. Commun., 15, 10352. Ovaskainen, O. & Saastamoinen, M. (2018). Frontiers in metapopulation biology: The legacy of Ilkka Hanski. Annu. Rev. Ecol. Evol. Syst., 49, 231–252. Palkovacs, E.P., Kinnison, M.T., Correa, C., Dalton, C.M. & Hendry, A.P. (2012). Fates beyond traits: Ecological consequences of human‐induced trait change. Evol. Appl., 5, 183–191. Purvis, A., Gittleman, J.L., Cowlishaw, G. & Mace, G.M. (2000). Predicting extinction risk in declining species. Proc. R. Soc. Lond. B Biol. Sci., 267, 1947–1952. Reed, J.M., Wolfe, B.E. & Romero, L.M. (2024). Is resilience a unifying concept for the biological sciences? Iscience . Reed, T.E., Schindler, D.E. & Waples, R.S. (2011). Interacting effects of phenotypic plasticity and evolution on population persistence in a changing climate. Conserv. Biol., 25, 56–63. Rocha, J.C., Peterson, G., Bodin, Ö. & Levin, S. (2018). Cascading regime shifts within and across scales. Science, 362, 1379–1383. Sjodin, B.M. & Russello, M.A. (2022). Comparative genomics reveals putative evidence for high-elevation adaptation in the American pika (Ochotona princeps). G3, 12, jkac241. Sjodin, B.M., Schmidt, D.A., Galbreath, K.E. & Russello, M.A. (2024). Putative climate adaptation in American pikas (Ochotona princeps) is associated with copy number variation across environmental gradients. Sci. Rep., 14, 8568. Smith, A.T. (2020). Conservation status of American pikas (Ochotona princeps). J. Mammal., 101, 1466–1488. Smith, A.T. & Weston, M.L. (1990). Ochotona princeps. Mamm. Species . Solari, K.A. & Hadly, E.A. (2020). Experimental study of hypoxia-induced changes in gene expression in an Asian pika, Ochotona dauurica. PLoS One, 15, e0240435. Stephenson, J.F. & Adelman, J.S. (2022). The behavior of infected hosts: behavioral tolerance, behavioral resilience, and their implications for behavioral competence. In: Animal behavior and parasitism (eds. Ezenwa, V.O., Altizer, S. & Hall, R.J.). Oxford University Press, Oxford, United Kingdom, pp. 287–302. Strauss, A., White, A. & Boots, M. (2012). Invading with biological weapons: the importance of disease-mediated invasions. Funct. Ecol., 1249–1261. Strauss, S.Y., Lau, J.A. & Carroll, S.P. (2006). Evolutionary responses of natives to introduced species: what do introductions tell us about natural communities? Ecol. Lett., 9, 357–374. Supp, S.R., Xiao, X., Ernest, S.K.M. & White, E.P. (2012). An experimental test of the response of macroecological patterns to altered species interactions. Ecology, 93, 2505–2511. Thorén, H. (2014). Resilience as a Unifying Concept. Int. Stud. Philos. Sci., 28, 303–324. Thorogood, R., Mustonen, V., Aleixo, A., Aphalo, P.J., Asiegbu, F.O., Cabeza, M., et al. (2023). Understanding and applying biological resilience, from genes to ecosystems. Npj Biodivers., 2, 16. Tribot, A.-S., Deter, J. & Mouquet, N. (2018). Integrating the aesthetic value of landscapes and biological diversity. Proc. R. Soc. B Biol. Sci., 285, 20180971. Varner, J. & Dearing, M.D. (2014). Dietary plasticity in pikas as a strategy for atypical resource landscapes. J. Mammal. Varner, J., Horns, J.J., Lambert, M.S., Westberg, E., Ruff, J.S., Wolfenberger, K., et al. (2016). Plastic pikas: behavioural flexibility in low-elevation pikas (Ochotona princeps). Behav. Processes, 125, 63–71. Vredenburg, V.T., Knapp, R.A., Tunstall, T.S. & Briggs, C.J. (2010). Dynamics of an emerging disease drive large-scale amphibian population extinctions. Proc. Natl. Acad. Sci., 107, 9689–9694. Winslow, E.M., Speare, K.E., Adam, T.C., Burkepile, D.E., Hench, J.L. & Lenihan, H.S. (2024). Corals survive severe bleaching event in refuges related to taxa, colony size, and water depth. Sci. Rep., 14, 9006. Figure 1. Conceptual model of relationships between resilience variables. We present a relatively simple, conceptual model encompassing many key resilience variables and their deterministic effects on each other. Black arrows denote aggregating from one biological scale to a larger one or from a full distribution to a metric (e.g., summing, averaging, etc.) while green arrows denote causal effects that do not involve aggregating. An individual organism (largest blue box) has a genotype that affects its phenotype (e.g., body size) and phenotype affects vital rates. Exogenous factors, such as the abiotic environment, can impact an individual’s phenotype via plasticity (dotted green arrow) or impact an individual’s vital rates (dashed green). Aggregating across the focal individual and other individuals (multiple blue boxes), we consider a population (largest orange circle) and various distributions in that population. Aggregating across a population distribution yields a population metric. Aggregating across populations of multiple species yields a community. The population of one species can be an exogenous factor altering the phenotype or vital rates of individuals of another species. Community distributions can be summed to yield certain community total metrics. Figure 2. Case study connections to key points. Colors denote different case studies with their disturbance and key taxa displayed as well. For each key point, the degree of support is listed and emphasized by text type with a brief explanation in the far right column. The degrees of support are labeled as Confirmed > Likely > Unknown > \soutNot applicable. Information & Authors Information Version history Copyright This work is licensed under a Non Exclusive No Reuse License.

Authors Metrics & Citations Metrics Article Usage 380views 252downloads Citations Download citation Jason Walsman, Mark Wilber, Luisa M Diele-Viegas, et al. The myth of resilience: Necessary non-resilience and the role of variable choice in ecosystem recovery. Authorea. 03 November 2025. DOI: https://doi.org/10.22541/au.176219659.90163271/v1 DOI: https://doi.org/10.22541/au.176219659.90163271/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu.