The importance of considering the duration of extreme temperatures when investigating responses to climate change

Abstract The frequency and duration of heatwaves are increasing because of human activities. To cope with the changes, species with longer generation times may have to rely on plastic responses. The probability that their responses are adaptive is higher if the species have experienced temperature fluctuations also in their evolutionary past. However, experimental studies investigating responses to heatwaves often use exposure times that are significantly shorter than recent heatwaves. We show that this can lead to faulty conclusions and that the duration of higher temperature has to be considered in experimental designs. We recorded the response of threespine stickleback to prolonged duration of higher temperature during the breeding season, using a population that has experienced large fluctuations in temperature in its past and, hence, is expected to endure temperature changes well. We found males to adaptively adjust their reproductive behaviours to short periods of higher temperature, but not to longer periods that extended across two breeding cycles. Males initially increased their reproductive activities—nest building, courtship and parental care—which ensured high reproductive success during the first breeding cycle, but decreased their reproductive activities during the second breeding cycle when exposed to sustained high temperature. This reduced their courtship success and resulted in fewer offspring. Thus, a species expected to cope well with higher temperature suffers fitness reductions when the duration of high temperature is prolonged. The results stress the importance of considering the duration of extreme environmental conditions when investigating the impact that human activities have on species. Responses to short‐term exposures cannot be extrapolated to assess responses to longer periods of extreme conditions.

species have experienced temperature fluctuations also in their evolutionary past.
However, experimental studies investigating responses to heatwaves often use exposure times that are significantly shorter than recent heatwaves. We show that this can lead to faulty conclusions and that the duration of higher temperature has to be considered in experimental designs. We recorded the response of threespine stickleback to prolonged duration of higher temperature during the breeding season, using a population that has experienced large fluctuations in temperature in its past and, hence, is expected to endure temperature changes well. We found males to adaptively adjust their reproductive behaviours to short periods of higher temperature, but not to longer periods that extended across two breeding cycles. Males initially increased their reproductive activities-nest building, courtship and parental care-which ensured high reproductive success during the first breeding cycle, but decreased their reproductive activities during the second breeding cycle when exposed to sustained high temperature. This reduced their courtship success and resulted in fewer offspring. Thus, a species expected to cope well with higher temperature suffers fitness reductions when the duration of high temperature is prolonged. The results stress the importance of considering the duration of extreme environmental conditions when investigating the impact that human activities have on species. Responses to shortterm exposures cannot be extrapolated to assess responses to longer periods of extreme conditions.

K E Y W O R D S
behaviour, climate change, courtship, global warming, heatwave, phenotypic plasticity, reproduction, spawning

| INTRODUC TI ON
Climate change is restructuring species communities across the globe (Antao et al., 2020). Assessing the mechanisms behind the changes and the vulnerability of individual species to the changes are important endeavours in our increasingly disturbed world. The knowledge is needed to develop strategies to mitigate large-scale changes to ecosystems, which otherwise could threaten their functioning and services (McLean et al., 2016;Urban et al., 2016). However, most experimental studies on the effects of climate change on species rely on short-term exposures to higher temperature. These brief exposures may not reflect the changes occurring in nature or expected to occur as climate change escalates (Noer et al., 2022). Thus, the degree to which existing results can be extrapolated to predict the vulnerability of species to the intensification of climate change is uncertain.
The vulnerability of a species to climate change depends on its ability to plastically adjust (including migration) and genetically adapt to the changes. Species with long generation times in relation to the rate at which the environment is changing may not be able to adapt through genetic modifications but have to rely on plastic adjustments (Barrett & Hendry, 2012;Chevin, 2013;Hendry et al., 2008;Merila & Hendry, 2014). These adjustments depend on existing reaction norms that have evolved under past environmental conditions (Chevin et al., 2010;Chevin & Hoffmann, 2017;Fox et al., 2019;Sih et al., 2011;Tuomainen & Candolin, 2011). Thus, the probability of adaptive responses is higher when the disturbance extends earlier encountered conditions, such as a gradual rise in temperature, than when it creates novel conditions (Candolin & Jensen, 2021;Sih et al., 2011). Nevertheless, various factors may still reduce the adaptiveness of the responses, such as the temporal pattern of the environmental change. In particular, prolonged periods of high temperature can pose a challenge to species that have only encountered shorter spells of higher temperature in their past (Angiletta, 2009;Roman-Palacios & Wiens, 2020;Stillman, 2019).
For instance, while individuals may cope well with brief heatwaves through reduced activity, the strategy may become maladaptive during extended heatwaves, as it may decrease foraging rate or the likelihood of finding partners (Candolin, 2019;Dell et al., 2014).
Temperature fluctuations are particularly troublesome for many ectothermic species, as their metabolism depends on the temperature of the external environment (Abram et al., 2017). Fluctuations that influence their metabolism can alter behaviours that determine fitness, such as reproductive behaviours (Brandt et al., 2018;Conrad et al., 2017;Macchiano et al., 2019). For instance, a short exposure to higher temperature alters courtship displays in the wolf spider Schizocosa floridana (Rosenthal & Elias, 2019) and in the field cricket Gryllus integer (Hedrick et al., 2002). However, the impact that prolonged duration of higher temperature has on reproductive behaviours is poorly known, as is the ultimate impact that the responses have on reproductive success (Bernal et al., 2020;Spinks et al., 2021;Suryan et al., 2021). Yet, the impact could alter population dynamics given the central role that reproductive behaviours play in determining fitness.
Species that reproduce in shallow coastal waters are particularly exposed to heatwaves, as the water warms up faster than offshore waters (Harvey et al., 2022;Oliver et al., 2021;Smale et al., 2019;Vinagre et al., 2018). The ability of these species to cope with sudden rises in temperature is likely to depend on their past experience of temperature fluctuations and, hence, on the presence of adaptive reaction norms for coping with fluctuations. However, past fluctuations may have been less intense, infrequent or of shorter duration than current ones and, hence, reaction norms based on past selection may not be adaptive under future climate change (La Sorte et al., 2021).
We investigated if a species that reproduces in shallow coastal waters-the threespine stickleback, Gasterosteus aculeatus-is able to adjust its reproductive behaviours to increases in water temperature, and if the ability depends on the duration of the higher temperature. Recent research indicates that exposure to higher temperature influences its reproductive success, but whether the impact depends on the duration of the higher temperature is unknown (Fuxjager et al., 2019;Wanzenboeck et al., 2022). We used a population that has experienced temperature changes in the past and which consequently is expected to cope well with brief periods of higher temperature, but whose ability to cope with longer periods is unknown. A recent heatwave in the study area in the summer of 2021 lasted for 50 days, being the warmest ever summer since the recordings started in 1845 (https://en.ilmat ietee nlait os.fi/opendata), and such conditions are expected to increase in the future (Meier et al., 2019). If a species that has been exposed to periods of higher temperature in the past is not able to adjust to the current increase in the frequency and duration of heatwaves, then species that originate from more stable environments may find it even more challenging. Thus, a failure of stickleback to adjust to prolonged periods of higher temperature could indicate an even stronger effect of climate change on other species.
To investigate the response of the studied threespine stickleback population to prolonged duration of higher temperature, and the impact that the responses have on reproductive success, we exposed males to higher temperature during either one or two breeding cycles. During reproduction, stickleback males build a tunnel-shaped nest out of algae to which they attract females to spawn using a courtship dance combined with nuptial coloration (Tuomainen & Candolin, 2013). Females leave after spawning and the male alone cares for the eggs in the nest until hatching. A male may complete several breeding cycles during one breeding season, usually two or three cycles (Candolin, 2000a). We recorded effects of the duration of higher temperature on reproductive behaviours-nest building, courtship, and parental care-and the ultimate impact the duration has on their reproductive success, the number of offspring produced.
We further assessed if the timing of the temperature increase-early or late during the breeding season-influences the responses. We predicted that the ability of males to maintain high reproductive activity would decline with the duration of higher temperature and that the effect would be more pronounced later in the season when males are in poorer condition (Candolin, 2000a).

| Collection and housing
We caught threespine stickleback in early May 2019 before the breeding season from a bay in the outer archipelago of the Northern Baltic Proper (60° N, 23° E) using Plexiglas traps (Candolin & Voigt, 2001) (see map in Supporting Information). Stickleback migrate to the bay from the open sea in the spring to spawn, and leave it at the end of the summer when the breeding season ends.
Temperature fluctuates in the bay across days (Granroth-Wilding & Candolin, 2022) and years (Meier et al., 2022), depending on climate conditions, because of its shallow topography; max depth is about 1.5 m. Thus, the population has experienced fluctuations in water temperature during the breeding season in its evolutionary past. We transported the fish to Tvärminne Zoological station, University of Helsinki, within 20 min. We housed the fish in large flow-through tanks (salinity 5.5 psu) at a density of 0.25 fish per litre, in an outdoor facility under natural temperature and light conditions. We fed the fish defrosted chironomid larvae once a day.

| Experimental design
When males came into reproductive condition, as determined by the development of nuptial colouration, we measured their size (standard length and weight) and transferred them to 10-L aerated tanks in climate chambers, one male per tank. We exposed them to one of four temperature treatments during two breeding cycles, 18 males per treatment: (1) constant normal temperature of 14°C, NN, (2) constant high temperature of 19°C, HH, (3) switch from high to normal temperature between breeding cycles, HN and (4) switch from normal to high temperature between cycles NH. The constant normal temperature served as the control and the constant high temperature as the long-term treatment, and the two treatments with switches as short-term treatments where the higher temperature occurred either early or later during the breeding season. A temperature of 14°C represents average natural temperature in the spawning habitat during the breeding season, as measured in June during 9 years preceding the study (14.1°C ± 0.7, mean ± SD, n = 9, see Table S1) (see also Granroth-Wilding and Candolin (2022) and MONICOAST-project of Tvärminne Zoological Station (www.helsi nki.fi/monic oast)). A temperature of 19°C or higher is occasionally attained when air temperature exceeds 25°C for several days.
However, during the preceding 9 years, temperatures of 19°C or higher had not lasted more than maximum 5 days (U. Candolin, unpublished data).
The duration of each treatment depended on the time it took a male to complete each breeding cycle, as males usually initiate the second cycle as soon as the first one is completed. Thus, males experiencing a switch in temperature were transferred to the other temperature treatment as soon as they had completed the first cycle.
Keeping the duration of the treatments equal across males would have forced us to prevent males from initiating the second breeding cycle-until the slowest male had completed his first cycle-which could have influenced their reproductive behaviour. To change water temperature, we transferred the tanks, containing the males, between the two climate chambers, allowing the water to gradually attain the new temperature, which took about a day. The tanks maintained at constant temperature were exposed to the same disturbance by moving them within the chambers. Males in the four treatments did not differ in body size (length F 3,68 = 0.73, p = .54, weight: F 3,68 = 1.73, p = .17, see Table S2 for values).
Each male tank contained a nesting dish (Ø 12.5 cm) filled with sand and filamentous algae, Cladophora glomerata, for nest construction, and an artificial plant for hiding (Candolin, 1997). LED lights above the tanks were programmed to replicate natural light conditions with dawn and dusk and a light:dark period that changed from 17:7 in May to 18:6 in June. To stimulate nest building, we presented the males with a gravid female, enclosed in a transparent, perforated, plastic jar, three times a day for 10 min. We recorded the time it took each male to build a complete nest, that is, a nest through which he had crept through (Candolin & Salesto, 2006).

| Courtship recordings
When a nest was ready, we recorded the courtship behaviour of the male towards three consecutive gravid females, with a 30 min break between presentations. Some females were reused among males, which was considered in the analyses. Each female was enclosed within a perforated, transparent jar and the presentation lasted for 10 min. Females had been kept in the outdoor facility under natural temperature conditions until 1-2 days before experimentation, when they were transferred to an 10-L aerated tank held at the same temperature as the experimental male tank. We video recorded the courtship behaviour of the male and analysed his behaviour using the software BORIS (Friard & Gamba, 2016): number of leads towards the nest (the male approached the female, sometimes through zigzag movements, and then moves in a straight line towards the nest), number of fanning bouts at the nest entrance (the male fans fresh water into the nest using his pectoral fins), total time spent fanning, and total time spent courting the female.

| Spawning and hatching success
After the third courtship recording, we released the female and allowed her to spawn with the male. We recorded the time until spawning occurred. If the female did not spawn within 2 h, we assumed she was not interested in the male and replaced her with a new female. Spawning time was in these cases noted as the maximum time, 120 min. We measured the number of eggs spawned by gently removing all eggs from the nest 2 h after spawning-when the eggs had hardened-and weighing the total clutch to the nearest 0.01 g (Candolin, 2000b). To calculate the number of eggs, we divided the weight of the egg clutch by the mean weight of an egg.
The weight of an egg was calculated by measuring the diameter of 10 eggs in the clutch using a camera connected to a cold-light microscope, and then multiplying mean egg volume with egg density (buoyance), 1.01 g/cm 3 (Nissling et al., 2017). Females spawn all ovulated eggs at one spawning. The number of eggs received did not differ among treatments (F 1,74 = 1.59, p = .21).
After weighing the egg clutch and photographing 10 eggs, which took maximum 10 min and does not influence survival (Candolin, 2000c), we returned the egg clutch to the nest. Males always accepted the eggs and resumed parental care behaviour. We recorded male parental behaviour by video recording the males for 10 min each day. We analysed the behaviours using the software BORIS: number of fanning bouts, total time spent fanning, and total time spent by the nest (which includes gluing, cleaning and prodding into the nest to remove dead or diseased eggs, in addition to fanning).
When the eggs were almost ready to hatch, after 7 days at 14°C and after 5 days at 19°C, based on earlier work , we removed the eggs from the nest and transferred them to separate aerated tanks to follow hatching success. We counted the number of fry emerging and calculated hatching success as the percent of eggs received that hatched. The possible presence of unfertilised eggs is unlikely to influence the results, as earlier studies show that average fertilisation success is over 99% in the population .
We allowed the males from the first breeding cycle to complete a second cycle by providing them with fresh material for nest building, and repeated the procedures from the first cycle. Four males did not enter a second cycle; one at normal temperature, and three at high temperature. In the analyses, qualitatively similar results are gained if these males are removed from the analyses or maintained as missing values during the second cycle. Given that we do not know the reason for the males not entering a second breeding cycle, we present the more conservative results with the males removed from the analyses. The total number of eggs received across the two breeding cycles did not differ among treatments (F 3,68 = 0.84, p = .48).
We measured the length and weight of the males at the end of the second breeding cycle and calculated changes in body condition as percent weight lost and as change in Fulton's condition factor K (Le Cren, 1951). Qualitatively similar results were gained using the two measures and we present the data for percent weight lost. We fed the males defrosted chironomids during the nest building stage, but not during courtship and parental care, as males do not feed during this period.

| Statistical analyses
We calculated principal components for the recorded courtship behaviours and for parental care behaviour (Tables S3 and S4). To assess if activity changed from the first to the second cycle depending on treatment, we used mixed models with treatment and cycle as fixed factors and male as random factor. To investigate the pattern in more detail, we separately tested if temperature influenced activity during the first cycle, using linear models with temperature as fixed factor, and, further, if activity changed from the first to the second cycle within each treatment, using mixed models with cycle as fixed factor and male as random factor. To consider the multiple use of some females, female identity was inserted into the models as a random factor. No significant effect of female identity was detected and we removed the factor from the models, which did not influence the results. All analyses were performed using the software IBM spss Statistics 26. We checked that the assumptions of the tests regarding heteroscedasticity and normal distribution of residuals were held.

| RE SULTS
Changes in activity from the first to the second cycle depended on treatment, as indicated by significant interactions between treatment and cycle for the recorded activities (Table 1). During the first breeding cycle, males reproducing at the higher temperature (19°C) were more active than males reproducing at the normal temperature (14°C) for the recorded behaviours: nest build- longer to build a nest and reduced their courtship and parental care activity ( Table 2). Males exposed to high temperature during the first cycle but not the second (HN) were slower at building a nest and reduced their courtship and parental care activity, while males exposed to high temperature during the second but not the first cycle (NH) showed the opposite pattern with faster nest building and increased courtship and parental care activity (

| DISCUSS ION
Our results show that threespine stickleback males adjust their reproductive behaviours-nest building, courtship and parental care-to changes in water temperature. They increase their activity when temperature rises and decrease it when temperature drops, as expected for ectothermic organisms (Dillon et al., 2010;Sinclair et al., 2016). However, when temperature continues to be high over a longer period-across two breeding cycles-males fail to maintain their high activity during the second cycle, which reduces their ability to attract females to spawn and to raise eggs in the nest to the hatching stage. Males that experience high temperature only during one breeding cycle (first or second), and who increase their activity only during this short period, are, on the other hand, able to maintain high reproductive success across the two breeding cycles.
Thus, enhanced reproductive activity during a short spell of high temperature is adaptive but not sustainable over longer periods.
Our expectation of sustained periods of higher temperature being more stressful than short periods was consequently upheld. These results show that a species that has been exposed to temperature fluctuations in its breeding habitat in the past, and is expected to withstand heatwaves better than many other species, fails to adjust to prolonged periods of high temperature and suffers a reduction in offspring production.
During the short spell of high temperature, males increased their reproductive activities irrespective of whether the spell occurred early or late during the breeding season. Thus, our expectation of males being less able to adjust to temperature increases later in the season, due to a gradual decrease in body condition (Candolin, 2000a), was not upheld. Given that higher temperatures are more common later in the breeding season, selection for adjusting to higher temperatures at this time of the year, also when condition starts to deteriorate, has probably been strong. Whether an effect of temperature would have emerged during a third cycle, as some males are able to complete three breeding cycles (Candolin, 2000a), remains to be determined.
The increased activity during a short spell of higher temperature was most likely a consequence of faster metabolic rate, as is generally the case in ectotherms (Dillon et al., 2010;Sinclair et al., 2016).
In addition, changes in female activity could have contributed to the enhanced courtship activity, as female metabolic rate also increases with temperature. Moreover, females could have been more willing to spawn at the higher temperature, as increased temperature accelerates the maturation of eggs and shortens the time window available for females to search for breeding males (Pankhurst & Munday, 2011).
Regarding the recorded increase in male parental care activity, this is required for successful rearing of eggs, as higher temperature lowers water oxygen saturation and accelerates the metabolic rate of eggs (Pankhurst & Munday, 2011;Smyder & Martin, 2002). Thus, the increased parental care activity during a short spell of higher temperature was adaptive, while the failure of males to maintain the high parental care activity during a longer period of high temperature apparently contributed to their lower hatching success.
The cause of the inability of males to maintain high reproductive activity under sustained high temperature-during the second breeding cycle-was probably exhaustion. This is supported by males losing more weight in the sustained high temperature treatment than in the other treatments. In addition, the nest building period between the two breeding cycles was prolonged for males on high temperature during the first cycle, which indicates the need for a longer recovery time. This prolongation partly cancelled out the benefit of a shorter parental care period, which otherwise could increase the number of breeding cycles that males could complete (Candolin, 2000a). Changes in female behaviour cannot explain the reduction in male activity and reproductive success during the second cycle (for males on sustained high temperature), as males exposed to high temperature only during the second cycle did not show a decline in activity or reproductive success (the females had experienced the same conditions as females in the sustained high temperature treatment). Similarly, a lower quality of eggs or fertilisation success cannot explain the lower hatching success of males on sustained high temperature (Mehlis & Bakker, 2014), as lower success was not recorded for males exposed to high temperature only during the second cycle.
TA B L E 2 Changes in reproductive behaviours and hatching success from the first to the second breeding cycle for threespine stickleback males in the four treatments: constant normal temperature (NN), constant high temperature (HH), high temperature during the first but not the second cycle (HN) and high temperature during the second but not the first cycle (NH). Note: N = 18 per treatment. Mixed models were used to analyse the data, with cycle as fixed factor and male as random factor.
These results indicate that threespine stickleback males are able to cope with shorter spells of higher temperature, but not when their duration is prolonged. Yet, such extreme conditions are becoming increasingly common. The investigated stickleback population has apparently adapted to short periods of high temperature, as these have occurred also in the past, but not to their prolongation. Thus, responses to short-term extreme conditions cannot be used to predict responses to longer-term changes for the investigated species. This stresses the importance of considering the duration of extreme conditions, such as heatwaves, when evaluating the ability of organisms to cope with the ongoing climate change. Investigations often focus on the limits of temperature tolerance, while effects of the duration of high temperature is seldom considered (Lawson et al., 2015;Rezende et al., 2014;Sinclair et al., 2016).
The results further stress the importance of considering the past history of species when predicting and assessing responses to extreme conditions (Chevin & Hoffmann, 2017;Sih et al., 2011;Tuomainen & Candolin, 2011). Although the experimental conditions only extended natural conditions, the population was unable to cope with their prolongation. The impact of prolonged periods of higher temperature could be even worse for species originating from more stable environments. Yet, little is currently known about the impact that past conditions have on the ability of species to cope with heatwaves (Bay et al., 2017;Candolin & Jensen, 2021;Lasky et al., 2020;Lavergne et al., 2013;Lawson et al., 2015;McGaughran et al., 2021).
To conclude, our results show that threespine stickleback males are able to adaptively adjust their reproductive behaviours to shortterm increases in temperature, but that the prolonged duration of high temperature poses a challenge that reduces the activity and reproductive success of stickleback. Thus, the average reaction norm that has evolved in the population is insufficient for coping with temperature rises when the duration of high temperature is prolonged.
These results emphasise the importance of considering the duration of extreme conditions when assessing the impact of climate change on species. Responses to short-term exposures cannot be extrapolated to assess responses to sustained periods of extreme conditions.

ACK N OWLED G M ENTS
We thank Tvärminne Zoological Station for providing working facilities and Swedish Cultural Foundation in Finland (grant number 150955 to UC) and Walter and Andrée de Nottbeck Foundation (to TI) for funding the work.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available in the Dryad Digital Repository (https://doi.org/10.5061/dryad.nzs7h 44tt) .