Conservation implications of ecological responses to extreme weather and climate events

Many conservation efforts now focus on mitigating biodiversity loss due to climate change. While a focus on impacts from mean, long‐term changes in climate is warranted, the vast majority of conservation plans largely ignore another key factor of climate change—changes in the frequency and intensity of extreme weather and climate events. A typology of the full range and severity of ecological responses to extreme events would help underpin tracking of their impacts.


| INTRODUC TI ON
Understanding what makes species and ecosystems vulnerable to climate change has become integral to the field of conservation science. To date, climate change vulnerability assessments have predominantly focused on how species and ecosystems will respond to mean, long-term changes in climate (Chapman et al., 2014;Jones, Watson, Possingham, & Klein, 2016), including regional warming (van Gils et al., 2016), seasonal shifts (Asch, 2015) and sea level rise (Runting, Wilson, & Rhodes, 2013). This focus has improved our overall understanding of how climate change is shaping the environment (Scheffers et al., 2016) but has left many conservation plans unprepared for changes in the frequency and distribution of extreme weather and climate events (Chapman et al., 2014;Jones et al., 2016). Moreover, it is increasingly recognized that changes in climate extremes can have greater environmental consequences than changes in climate means (Bailey & van de Pol, 2016;Gutschick & BassiriRad, 2003;Harris et al., 2018;Maron, McAlpine, Watson, Maxwell, & Barnard, 2015;Vasseur et al., 2014).
Extreme weather and climate events (weather or climate events that are rare within their statistical reference distributions at a particular place (IPCC, 2014); herein "extreme events"), such as cyclones, floods, heat waves and drought, have become more frequent and intense in many regions of the world as a consequence of anthropogenic climate change (Seneviratne et al., 2012;Ummenhofer & Meehl, 2017). This pattern is likely to accelerate. For example, increases in daily temperature and precipitation extremes are likely to continue, more extreme rainfall is expected in southern Africa, increased drought intensity is expected in central America, northeast Brazil and the Mediterranean, and drought and heat waves are expected to become more frequent in Australia, northern Africa and south-western America (Handmer et al., 2012;Pohl, Macron, & Monerie, 2017;Russo, Marchese, Sillmann, & Imme, 2016;Seneviratne et al., 2012).
Our knowledge of ecological responses to extreme climate stress is predominantly based on laboratory or field experiments (Bailey & van de Pol, 2016;Hoffmann & Parsons, 1997). Fundamental ecological insights gleaned from such studies, including thermal tolerance of a species or tipping point for an ecosystem, can potentially aid in the design of effective conservation interventions. However, the utility of experimental studies for conservation intervention design is limited, because experimental studies rarely replicate the variable and highly nuanced bioclimatic conditions species face in unmodified ecological systems (Chevin & Hoffmann, 2017).
Observational studies of species' or ecosystems' responses to extreme events can highlight particular events or taxonomic groups that may benefit from focused vulnerability research (Altwegg, Visser, Bailey, & Erni, 2017;Smith, 2011). Some dramatic ecological responses to extreme events have been observed across individual, population and ecosystem scales. For example, cyclones can alter the onset of sexual maturity in turtles (Dodd & Dreslik, 2008), prolonged droughts have caused population collapse in koalas (Seabrook et al., 2011), flooding can reduce plant species richness (Miller, Gornish, & Buckley, 2010), and heat waves can alter the structure of marine ecosystems (Wernberg et al., 2013).
Previous reviews of observed responses to extreme events in ecological systems have focused on a specific taxonomic group (Orsenigo, Mondoni, Rossi, & Abeli, 2014;van de Pol et al., 2010), extreme event (Cech & Cech, 2013;Lugo, 2008), or set of responses (Moreno & Moller, 2011), or indicate potential responses to extreme events without quantifying the magnitude of the impacts over time (Bailey & van de Pol, 2016;Easterling et al., 2000;Jiguet, Brotons, & Devictor, 2011;Parmesan, Root, & Willig, 2000). To our knowledge, there has been no formal review and categorization of the full range of observed responses to extreme events in ecological systems, which could be used to inform effective conservation strategies.
Here, we categorize and quantify observed ecological responses to extreme events. We build on the existing literature by reviewing studies spanning 70 years of how seven taxonomic groups (birds, mammals, fish, amphibians, reptiles, invertebrates and plants) have responded to five types of extreme event (drought, cyclones/hurricanes/typhoons, floods/storm surges, heat waves, and cold waves/ extreme winters/ice storms). We also collected information on study duration and whether studies report on species or system recovery following exposure to an extreme event. We discuss future research needs and conceptual challenges associated with incorporating extreme events into conservation vulnerability assessments and adaptation plans.

| ME THODS
We performed a literature search in the ISI Web of Science database in July, 2015 and included all literature published prior to that date.
Key search terms were set according to a "taxon" AND "event" structure and changed iteratively until all combinations of taxa and event included in our review were searched. Additional search terms were used to better refine search results (see Supporting Information   Table S1). Search results (n = 4,896) were screened for our inclusion criteria: (a) an observational study; (b) from peer-reviewed literature published in a national or international journal; and (c) documenting an ecological response during, or in the years following an extreme event. Studies reporting on responses to seasonal flooding (e.g., in the Brazilian Pantanal) or non-climate-induced flooding events (e.g., dam decommission) were excluded, as were studies of commercially farmed or produced species.
We read studies that met our inclusion criteria (n = 584), and recorded the documented ecological response, the focal species (if applicable), the ecosystem in which the response was studied, whether responses were observed across multiple years, and whether studies were designed to monitor species or ecosystem recovery following exposure to an extreme event. After initial review, 65 studies were excluded because they did not meet the inclusion criteria, which left 519 studies for review (see Supporting Information Table S2  Similar ecological responses to extreme events were grouped together and categorized (Table 1). Responses were further classified as being positive, neutral, negative or ambiguous, where ambiguous responses were those that could imply adaptation or maladaptation to an extreme event. Extreme events with similar definitions were grouped into a single category (i.e., cyclones, hurricanes and typhoons; floods and storm surges; and cold waves, extreme winters and ice storms). Many studies documented more than one ecological response. As such, our review includes more responses (n = 698) than studies (n = 519; Figure 1). Decline in estimated population abundance of less than 25%, or any unspecified per cent loss in population abundance Population decline >25% Decline in estimated population abundance of greater than 25% Little impact Rapid recovery or no response following exposure to extreme event

| RE SULTS
Our review revealed nineteen different types of responses (hereafter, response categories) ( Table 2; Supporting Information Table   S3). Four of these response categories were positive responses, six were ambiguous, eight were negative, and one category included cases where little or no response to an extreme event was observed.
Plants were the most commonly studied taxonomic group (number of responses, n = 189), followed by birds (143) Table S4).
Sixty per cent of the studies in our review observed ecological responses for more than one year (n = 416). Studies that focused on drought, cold waves and floods generally spanned more than 1 year, whereas more than half of cyclone and heat wave studies were completed within 1 year (Supporting Information Table S5). We found no discernible differences in the type or frequency of responses to extreme events observed from studies carried out within 1 year compared to studies carried out across multiple years (Supporting Information Table S6).
Only 40% of studies monitored species or ecosystem recovery following exposure to an extreme event (Supporting Information   Table S7). Of these studies, 38% showed species or ecosystems that did not recover to pre-disturbance levels after extreme event exposure, while full or partial recovery was observed in 62% of cases.
Recovery usually occurred within 2 years, but in some cases, species or ecosystems took over 10 years to recover fully to pre-disturbance levels (Supporting Information Table S7).
Negative ecological responses were the most commonly reported, accounting for 57% of all documented responses (n = 395;  Positive responses to extreme events were the least frequent of all impact categories (n = 68). Over half of these positive responses were population increases (n = 39), which primarily followed a cyclone or flood event. There were very few positive responses to cold waves (n = 2) and heat waves (n = 1), and no positive reptilian responses to any extreme event.
Cyclones were the most prevalent extreme event for birds, fish, plants and reptiles. Among these taxa, reptiles seem to be particularly responsive to cyclones (n = 14, 52% of all reptile responses to extreme events), with the majority of these responses being negative (n = 11, 79%). Cyclones also led to 15 cases of a > 25% population decline in bird populations (58% of all negative bird responses to cyclones).
There were 27 cases of plants declining in body condition following a cyclone, with eight of these responses persisting long after the event.

| D ISCUSS I ON
Collectively, the studies in our review suggest that extreme weather and climate events have profound implications for species and ecosystem management. Ecological responses were observed across all ecological levels, from individuals to ecosystems, and over half were found to be negative. Some of the more severe negative responses identified in our review were prevalent and persisted long after an extreme event. There were, for example, over 100 cases of a >25% population decline and 31 cases of local extirpation of a species. The numerous declines in fecundity and species richness documented further justify the inclusion of extreme event impacts in conservation planning frameworks.
Plant species had the highest proportion of negative responses to extreme events (70%). Many of these involved structural damage (e.g., uprooting, loss of stem density, increased herbivory) following cyclones-which in many cases would be impermanent-but there were also numerous cases of plant communities suffering increased mortality rates and reduced species diversity following a cyclone.
Reptile and amphibian species also had a high proportion of negative responses to extreme events (both 67%). Declines in fecundity after cyclones and reduced body condition following drought were prevalent in reptilian species, while populations of amphibian species suffered large declines following cyclone events. Species located in coastal ecosystems had the highest proportion of negative responses (65%), which included fecundity declines and population declines of >25% following cyclones. Large population declines following drought events were also numerous in terrestrial ecosystems.
Many of the studies we reviewed observed ecological responses for <1  While widespread ecological changes following extreme events may prove to be lasting or even irreversible (Harris et al., 2018;Pisaric et al., 2011), long-term observational studies also provide valuable insights into how ecosystems can recover following an extreme event (Capon et al., 2015). For example, near Glacier Bay, Alaska, the density of pink salmon (Oncorhynchus gorbuscha) decreased by 90% following a major rainfall event that caused substantial geomorphic change to stream channels. Within two generations, however, salmon densities recovered to pre-disturbance levels (Milner, Robertson, McDermott, Klaar, & Brown, 2013). Bird communities in far north Queensland, Australia can also recover to a pre-disturbance state within 12 months of a tropical cyclone by modifying their foraging behaviour and movement patterns within a landscape (Freeman, Pias, & Vinson, 2008). Hence, the markedly different responses to extreme events observed in long-term studies show that caution is required when predicting likely responses to future events. The information gleaned from long-term studies will be critical for improving predictions of species responses to extreme events; however, such studies are rare.
We found a larger-than-expected number of positive or neutral responses to extreme events (n = 144; 21% of all responses).
These responses serve as a reminder that natural disturbances from extreme events often play a critical role in maintaining the structure and function of many ecosystems and life history strategies (Attiwill, 1994;McMahon et al., 2017;Whittaker, Willis, & Field, 2001). Some positive responses that were common yet temporary included cyclone events leading to increased abundances of bird species that prefer regenerating habitat (e.g., Brown, Sherry, & Harris, 2011;Freeman et al., 2008), and cyclones and floods leading to increased richness of fish and invertebrate species (e.g., Gerisch, Dziock, In the studies we reviewed, species that benefited from extreme events were typically invasive (e.g., green sunfish (Lepomis cyanellus) in California ( (Terrell, Engbrecht, Pessier, & Lannoo, 2014). In these cases, cyclones and drought conditions act to create microhabitat conditions that exceed the critical maximum temperature for chytrid fungus and hence help anuran resistance to the disease.

| Assessing vulnerability to extreme events
Vulnerability assessments can identify species or ecosystems that are likely to require conservation intervention. A widely adopted framework for assessing vulnerability to climate change that measures species' exposure, sensitivity and adaptive capacity (Foden et al., 2013) is also applicable to measurements of species' vulnerability to extreme events. Exposure is a measure of the nature and degree to which a system is exposed to significant climatic variations.
Sensitivity is a measure of how a species will be affected by particular changes in climate variables, and is predominantly governed by intrinsic factors (e.g., phenology, physiological traits). Adaptive capacity is the ability of a species to adjust to altered climate and weather patterns, and is influenced by both intrinsic (e.g., strict habitat dependencies) and extrinsic factors (e.g., habitat loss waves are likely to be more intense, more frequent and last longer, particularly in Amazon and Congo basins (Seneviratne et al., 2012).
The frequency of tropical cyclones may remain unchanged but they are predicted to become more intense in some ocean basins (Seneviratne et al., 2012). Moreover, the duration and intensity of droughts will increase in some regions of the world, including southern and central Europe, central North America, Central America and Mexico, north-east Brazil and southern Africa (Handmer et al., 2012;Pohl et al., 2017;Russo et al., 2016;Seneviratne et al., 2012).
The infrequent and short-term nature of extreme events makes it difficult to predict their local frequency and intensity (Bailey & van de Pol, 2016;Flato et al., 2013). Predictions are also constrained by incomplete data on the historical frequency of extreme events for some regions (Flato et al., 2013). Of the extreme events included in this review, predictive modelling is least developed for heat waves that affect marine ecosystems (Frölicher, & Laufkötter, 2018), whereas predictions for extreme precipitation and heat events have improved dramatically in recent years (e.g., Baker et al., 2018;Tabari & Willems, 2018 (Beehner et al. 2006). But despite a long history of experimental research on ecological responses to climatic stress Hoffmann & Parsons, 1997;Pardo, Jenouvrier, Weimerskirch, & Barbraud, 2017;Parmesan, 2006;White, Campbell, Kemp, & Hunt, 2000), key questions on species' sensitivity and adaptive capacity to extreme events remain. Such questions include those around limits to phenotypic plasticity, the capacity for genetic change in individuals, and how individual responses to extreme events influence ecosystem function (Palmer et al., 2017). Sessile species with slow reproductive rates and specialized habitat and dietary requirements are commonly flagged as being highly vulnerable to anthropogenic climate change (Pacifici et al., 2015). These same traits are also likely to exacerbate vulnerability to more frequent and intense extreme events.
However, research that continues to explore this important research gap will help clarify the preconditions for successful adaptation to extreme events among different taxonomic groups.
Studies exploring species' sensitivity and adaptive capacity to extreme events will inevitably require large investments of both money and time, and hence should target areas of most need. Our findings suggest priorities for this research include ecological responses to cold and heat waves, given their predominantly negative impacts (73% and 74% of responses were negative, respectively) and the comparatively few studies focusing on these events. Responses in coastal, estuarine, marine, riparian and wetland ecosystems were mostly negative and relatively poorly studied (Figure 3; Supporting Information Table S8). Other less-studied topics include reptile responses to cyclones, and mammal and bird responses to drought.
Future studies should also examine ambiguous responses, including changes in behaviour and species composition, to help clarify their long-term implications. Changes in the composition of invertebrate communities following drought and flood events, and changes in avian species behaviour following cyclones and floods were particularly prominent.

| Accounting for threat interactions in vulnerability assessments
Understanding how multiple threats interact to influence species' vulnerability to extinction is difficult (Cote, Darling, & Brown, 2016), but can be of critical importance to the success of conservation efforts (Brook, Sodhi, & Bradshaw, 2008). Assessments of species' vulnerability to extreme events should consider how interactions between threats may exacerbate or attenuate vulnerability levels. Such interactions include those between multiple extreme events or between extreme events and other threatening processes.
Multiple extreme events are likely to act in synergistic ways to exacerbate risk of species' extinction. For example, the co-occurrence of drought and heat waves has greater impacts on bird abundance changes than if these events occur in isolation (Albright et al., 2010). The combination of heat waves and low summer rainfall also has severe impacts on koala populations (Phascolarctos cinereus), either directly by causing physiological stress or indirectly by affecting the nutrient and water content in eucalypt leaves (Seabrook et al., 2011). Synergistic interactions between extreme events have also been documented in marine systems. For example, heat waves increase demand for carbon in a temperate seagrass species (Amphibolis antarctica) found in Shark Bay, Western Australia. Yet, this demand cannot be met through photosynthesis when turbid floodwaters reduce light availability, resulting in a negative carbon balance in plants for more than 2 years after the co-occurrence of heat waves and flooding events (Fraser et al., 2014).
Interactions with non-climate-related threats are also likely to influence species' overall vulnerability to extreme events. For example, habitat fragmentation can limit the ability of butterflies to cope with, and recover from, heat waves (Piessens, Adriaens, Jacquemyn, & Honnay, 2009). Toxic compounds in agricultural runoff are also less diluted in years of drought, which can elevate toxicity levels in fish liver tissue (e.g., in striped bass (Morone saxatilis) in the San Francisco Bay estuary (Bennet, Ostrach, & Hinton, 1995)). Drought can also promote a complex interaction between diseases that are tolerable in isolation, but with co-infection cause catastrophic mortality in lion populations (Panthera leo) (Munson et al., 2008). Sustained browsing by introduced ungulates during drought events has been linked to declines of a critically endangered seed specialist bird (Loxioides bailleui) in Hawaii (Banko et al., 2013).
Ecological responses are likely to be more pronounced when extreme events co-occur with other events or threatening processes, particularly where species exist near upper thermal tolerance limits (Fraser et al., 2014), are resource-limited (Maron et al., 2015) or have specialized habitat requirements (Banko et al., 2013;Hinojosa-Huerta, Nagler, Carrillo-Guererro, & Glenn, 2013). Accounting for threat interactions is likely to improve predictions of responses to extreme events, such as for population decline (Vasseur et al., 2014).
Further research that reviews or elucidates when extreme events act additively, synergistically or antagonistically with other threats will therefore improve the reliability of species vulnerability assessments. To advance on current knowledge of threat interactions, such research should be long term and include measures of threat intensity. However, we caution that research focusing on interactive effects should not detract from efforts to clarify how species are likely to respond to altered extreme event patterns, which remains a key uncertainty in most vulnerability assessments.

| Promoting adaptation to extreme events through conservation action
Just as they promote adaptation to climate change, actions that enhance habitat connectivity, access to climate refugia and intra-species genetic variation are also likely to benefit species threatened by extreme events. The most effective way to achieve these aims is through the conservation of intact habitats (Martin & Watson, 2016;Watson et al., 2018).
Intact habitats on land are typified by large, contiguous areas of native vegetation that often span environmental gradients, such as altitude, rainfall or temperature . Marine intact habitats (Jones et al., 2018) are seascapes mostly free of human disturbance that perform key functional roles (D'agata, 2016) and maintain high levels of ecological and evolutionary connectivity (Jones, Srinivasa, & Almany, 2007 (Alila, Kuraś, Schnorbus, & Hudson, 2009;Brookhuis & Hein, 2016). Intact habitats are also likely to be more resilient to large-scale disturbances from extreme events. For example, exotic seedling germination following cyclone damage is higher and more diverse inside fragmented habitats than intact habitats (Catterall, McKenna, Kanowski, & Piper, 2008).
Where intact habitat protection is not available to conservation practitioners, ecological restoration efforts can also help species to adapt to extreme events (Reside, Butt, & Adams, 2017). For example, sustained restoration efforts (i.e., 15+ years) in brackish marshes help plant and animal communities to cope with drought events (Kinney, Quigg, & Armitage, 2014). Ecological restoration that helps species to adapt to extreme events can also benefit human communities with immediate adaptation needs (Maxwell, Venter, Jones, & Watson, 2015). There are now important examples of using oyster or seagrass beds to protect coastal areas from flooding offers substantial dual benefits for climate-vulnerable biodiversity and human communities (Borsje et al., 2011).
In fragmented landscapes, re-establishing native vegetation and regulating incompatible land uses will facilitate species mobility during extreme events and provide critical food and shelter resources to aid recovery after events (e.g., Steenhof, Kochert, Carpenter, & Lehman, 1999). Populations that face resource bottlenecks during or following extreme events would benefit from feed supplement programmes (Maron et al., 2015;Turton, 2012) or removing competition from invasive ungulates (Banko et al., 2013). It may not be possible, however, to restore ecosystem stability and community assemblages to better cope with extreme events and other stressors in areas that have been heavily degraded (Cardoso et al., 2008).
Restoring environmental flows or improving management of groundwater withdrawals during drought periods will become increasingly important to maintain many freshwater populations and systems threatened by drought and heat wave events (Baker & Jennings, 2005). Importantly, the success of such restoration efforts will depend on also re-establishing natural temperature regimes in river and stream systems-an effect that is difficult to achieve if water is simply released from dams or reservoirs at periodic intervals (Rader, Voelz, & Ward, 2008).
More intensive conservation interventions could be necessary for critically endangered species that are vulnerable to extreme events.
Such actions could involve pre-emptive translocation or relocation of populations following successive extreme flood events (e.g., Sousa et al., 2012). Populations that are regularly exposed to extreme events, or inhabit highly variable environments, may be suitable source populations for ex-situ conservation efforts as such populations typically show higher phenotypic plasticity and may be preadapted to more frequent and intense extreme events (Chevin & Hoffmann, 2017).

| Concluding remarks
Our review provides an overview of contemporary ecological responses to extreme events, and lays a foundation for future long-term studies to improve the understanding of species sensitivity and adaptive capacity to extreme events. Predicting the occurrence of individual extreme events and subsequent ecological responses is likely to remain a challenge. Less-documented phenomena include ecological responses to heat and cold waves, reptilian responses to cyclones, mammalian and bird responses to drought, and clarifying ambiguous ecological responses. Incorporating extreme events into climate change vulnerability assessments and adaptation plans will be challenging, but by doing so we have a greater chance of arriving at conservation interventions that truly address the full range of climate change impacts.

ACK N OWLED G EM ENTS
We

DATA ACCE SS I B I LIT Y
The studies included in this review are listed Supporting Information