Reproductive phenology and the influence of temperature in two sympatric New Zealand freshwater mussel species

Brooding status assessments

At all sites, E. menziesii and E. aucklandica sampling was undertaken fortnightly during the late austral spring–summer (excluding December because of high flow events) when female larval brooding activity is known to increase, and monthly at other times over a continuous period of 13 months. Tactile and visual (bathyscope) searches were conducted along the same reaches (15–40 m) on each occasion until 20–40 individuals of each species were collected. To determine their length, all individuals were measured to the nearest millimetre by using calipers. Additionally, the sex of each individual was determined by observing the gill morphology, specifically looking for marsupiumabsence (male) or presence (female) by using a speculum to prise apart valves. In sexually mature female hyriids, the marsupium is clearly visible throughout the year on the inner demibranchs as opaque thickened columns perpendicular to the hinge of the shell (Jones et al. 1986; Jupiter and Byrne 1997; Walker et al. 2001; see Fig. 2). Female brooding status was assessed using the procedure outlined by Melchior et al. (2021; see also Fig. 3), on the basis of classification of gill volume and pigmentation. During this process, individuals were held at ambient water temperatures in 2.5-L buckets or in mesh nets within the stream. Individual mussels were returned to their original location at the end of each sampling event.

Fig. 2.

(a) Non-brooding Echyridella aucklandica female with empty marsupium clearly visible on ventral inner demibranch (insert), and (b) male E. aucklandica with no marsupium present. (c) Mussels shown are from Kahururu Stream, with E. aucklandica (left) and E. menziesii (right) showing differences in size.

Fig. 3.

Larval stage microphotography, gill pigmentation gradient and description for staging validation in Echyridella aucklandica and E. menziesii (adapted from Melchior et al. 2021).

Six additional individuals per species collected upstream of each sampling reach during each sampling event (except March 2018) were used for validation of in vivo brooding stage by extraction of eggs or glochidia from gill marsupia (Fig. 3). The extracted gill contents were examined in the laboratory on a Petri dish under an Olympus SZ-6045 dissecting microscope (20–40×) and characterised into one of the five developmental stages (Fig. 3).

Environmental parameters

Water temperature (°C) was recorded at 15-min intervals at all sites for the duration of the study by using temperature loggers (Hobo Tidbit Data Logger, Onset, Bourne, MA, USA) submerged close to the streambed. Accumulated degree-days (ADD hereafter) were calculated with the University of California’s State-wide Integrated Pest Management Program online degree-day calculator (see https://ipm.ucanr.edu/WEATHER/index.html#DEGREEDAYS; Baskerville and Emin 1969), which uses maximum and minimum daily temperatures to calculate the number of ADD from set dates on the basis of the single sine method (Higley et al. 1986). Start dates were set from (i) the end of the previous brooding period for E. aucklandica and E. menziesii (31 March 2018) to determine ADD required to reach the next season’s brooding onset, and (ii) the sample date at which brooding mussels were first observed for each species within each stream (see Results) to determine the ADD required to reach peak brooding proportion as well as the duration of the brooding period.

The minimum thermal limit used for calculating ADD for each species was defined as 10°C, on the basis of laboratory trials (see the ‘Context’, ‘Methods’ and ‘Results’ sections of the Supplementary material) that tested for low thermal tolerance in brooding E. aucklandica and E. menziesii females. Here, the endpoint was defined as the temperature at which reproduction was likely to be halted and larvae prematurely aborted. The minimum threshold temperature used in this study corresponds to that found in similar international studies of unionids of the genera Quadrula, Cyclonaias and Fucsonia for ADD calculations based on metabolic-rate data (Spooner and Vaughn 2008; Galbraith and Vaughn 2009; Dudding et al. 2019).

Statistical analyses

All analyses were performed using the software R (ver. 4.0.0, R Foundation for Statistical Computing, Vienna, Austria, see https://www.r-project.org/). Data were assessed for normality (Shapiro–Wilk’s test and Q–Q plots) and homogeneity of variances (Levene’s test), and because parametric assumptions could not be confirmed for univariate statistics, non-parametric tests were used, with statistical significance defined at P ≤ 0.05. Following on from this, two-way Kolmogorov–Smirnov two-sample goodness-of-fit tests were used to detect differences in cumulative distribution of brooding proportions over time between each species within each site. Kruskall–Wallis tests (followed by Dunn’s post hoc multiple comparisons tests with Bonferroni correction) were performed to test for differences among sites in ADD required to reach brooding onset, peak brooding, and duration of brooding for each species.

To determine the strength of relationships between water temperature, specifically thermal summation (i.e. ADD), and brooding onset and brooding peak in E. menziesii and E. aucklandica, beta regressions with logit link function were implemented using the betareg package (ver. 3.1-3, A. Zeileis, F. Cribari-Neto and B. Gruen, see https://cran.r-project.org/package=betareg/; Cribari-Neto and Zeileis 2010), with ADD as the single predictor variable and the following response variables analysed separately for E. menziesii and E. aucklandica on the basis of pooled data from multiple sites: (i) proportion of mussels with glochidia development at Stage 1 (i.e. brooding onset), and (ii) proportion of mussels with glochidia development at Stage 4 (peak brooding). The coefficients of determination (r²) calculated using beta regression with logit function are given for each model. Beta regression was used to analyse these data, instead of a Gaussian generalised linear model, because the response variables (brooding onset and brooding peak) were expressed as proportions with a range from 0 to 1 and had variances that are usually not constant across the range of the predictor (Douma and Weedon 2019). Beta regressions can account for the heteroskedasticity and skewness that are common with proportional data (Cribari-Neto and Zeileis 2010). To help interpret the magnitude of the effects of the beta regressions, average marginal effects (which measure the change in a response given a change in a covariate) were calculated for each model by using the margins package (ver. 0.3.26, T. Leeper, J. Arnold, V. Arel-Bundock and J. A. Long, see https://cran.r-project.org/package=margins/; Williams 2012). Model assumptions, including normality and heteroskedasticity of residuals, were verified and assessed visually by diagnostic plots.

Stream temperature regimes

Ohautira had the lowest average temperature every month throughout the entire sampling period (Table S1 of the Supplementary material), reaching a minimum from May to June of 6.4°C (Ohautira), compared with 7.1°C at Mangapiko. Warmest temperatures were experienced in February when averages varied between 18.9°C (Ohautira) and 20.2°C (Kahururu), with maxima at Kahururu of 24.2°C, followed closely by Pakoka (24.1°C), and then Mangapiko (22.9°C) and Ohautira (22.1°C) (Table S1). Lowest average February temperatures ranged from 9.8°C (Ohautira) to 10.9°C (Mangapiko) (Table S1).

Mussel population characteristics

At all sites, valve lengths were skewed towards larger mussels, particularly for E. aucklandica, with few mussels below 40 mm (Table S2). Gravid females were identified across a range of size classes within each species. The smallest brooding female found in this study was 23 mm long for E. menziesii (Pakoka) and 40 mm for E. aucklandica (Mangapiko), whereas the largest brooding E. aucklandica collected was 110 mm (Ohautira), compared with 91 mm for E. menziesii (Pakoka) (Table S2).

In total, 1813 assessments of female brooding status were conducted across all sites, species and dates combined. Brooding season varied among species, with brooding across the four sites starting 1–3 months earlier in E. aucklandica (austral autumn) than E. menziesii, which initiated brooding in late winter–early spring. In E. aucklandica, the timing of brooding onset varied among sites, ranging from late May to mid-July, unlike in E. menziesii, which consistently began brooding larvae at all sites in August when water temperatures started to show steady increases (Fig. 4, 5).

Fig. 4.

(a–d; upper panels) Brooding proportions in female Echyridella aucklandica and E. menziesii over 1 year (March 2018–March 2019) at the four sampling sites (see Fig. 5 for detail). (a–d; lower panels) Monthly temperature plots are shown for comparison; points indicate means and vertical lines indicate 10th and 90th percentiles (see Table S1 for detail).

Fig. 5.

Glochidia brooding stages (1–5; see Fig. 3) over time (March 2018 to March 2019) in female Echyridella aucklandica (left) and E. menziesii (right) at four sites.

Brooding proportions at all sites except Mangapiko increased gradually over time, with females of E. aucklandica and E. menziesii reaching ~90% gravidity at peak brooding (Stage 4) within each site (Fig. 5). Females of E. menziesii and E. aucklandica were observed to have unimodal brooding peaks, with only one peak reproductive event at each site. Peak brooding (Stage 4) occurred slightly earlier in October–November (spring) and remained high for longer in E. aucklandica than in E. menziesii, which reached maximum brooding proportions in November and December (late spring to early summer) (Fig. 4). Gradual declines in brooding proportions following the peak at each site suggest the start of glochidia release in both species, with minimum brooding proportions in February to March (mid-summer to autumn) for E. aucklandica (coinciding with decreasing water temperatures) and January to February (summer) for E. menziesii, coinciding with maximum stream temperatures (Fig. 4).

Within all sites, the overall brooding season for female E. menziesii was shorter (6–7 months) than that for female E. aucklandica, which remained gravid for 9–11 months (Fig. 4). Although brooding for E. aucklandica occurred earlier and continued for longer than that for E. menziesii, no statistically significant variations were detected in the overall brooding distributions between the species at each site (Ohautira: D = 0.3, P = 0.7; Kahururu: D = 0.2; P = 0.99; Pakoka: D = 0.3 P = 0.6, Mangapiko: D = 0.3, P = 0.6), in part owing to temporal overlaps among sites in peak and non-brooding periods.

Brooding stages determined in vivo (verified by in vitro gill examinations) found that glochidia development within both species across all sites was initially synchronous. Early embryos made up the highest proportions of developing glochidia for several months following brooding onset. These were progressively replaced with later developmental stages (Fig. 5). Throughout mid-brooding, various developmental stages were present within individuals of both species throughout each sample month, until the period preceding the final stage of glochidia maturation (Stage 4). Mature glochidia (Stage 4) within E. aucklandica marsupia first occurred in September or October at all sites, whereas mature glochidia in E. menziesii marsupia were observed 1 month later in October and November across sites (Fig. 5).

Effects of temperature on brooding phenology

Analysis of ADD showed variable thermal regimes experienced by E. aucklandica and E. menziesii over time among sites (Fig. 6). The general observed upward trend in ADD steadied throughout the cooler months (May–July) and increased from August and September onward when water temperatures began to increase (Fig. 6). Despite being widely separated geographically, Kahururu (coastal) and Mangapiko (inland) had synchronous ADD regimes throughout almost the entire sample year and experienced similar increasing rates in ADD from October to November onward. Pakoka, too, followed patterns similar to those of Kahururu and Mangapiko, but experienced a slowing in ADD from November, whereas Ohautira’s forested headwaters and riparian shade meant that the ADD curve was lower throughout the entire study period than for the other sites (Fig. 6).

Fig. 6.

Cumulative stream temperature curves as degree days for each stream, showing phenologically sensitive time periods: EA onset, first observation of brooding in Echyridella aucklandica (solid vertical line) at each site; EM onset, first observation of brooding in E. menziesii (dotted vertical line); peak brooding, month at which brooding proportions were highest for each species.

Significant differences between E. aucklandica and E. menziesii in median ADD required to reach brooding onset were found across all sites (Ohautira: P = 0.018, Kahururu: P = 0.01, Pakoka: P < 0.001, Mangapiko: P < 0.001; Table S3 of the Supplementary material). First observations of brooding in E. aucklandica varied substantially in terms of ADD (321, 326, 412 and 426 ADD at Ohautira, Pakoka, Mangapiko and Kahururu respectively; Table 2). Median ADD required to reach brooding Stage 1 for each E. aucklandica varied significantly among sites (K–W: n = 81, H = 424.2, P = 0.04), with differences occurring between Ohautira (x~ = 431 ADD) and Pakoka (x~ = 471 ADD adj P = 0.03), and between Ohautira and Kahururu (x~ = 494 ADD, adj P = 0.05), but not between the other sites. Echyridella menziesii showed a similar wide range of ADD among sites for brooding onset (345, 448, 454, and 484 ADD at Ohautira, Kahururu, Pakoka and Mangapiko respectively) (Table 2). Median ADD required to reach Stage 1 for each E. menziesii varied significantly among sites (K–W: n = 70, H = 30.9, P < 0.001), with Ohautira (x~ = 431 ADD) requiring lower ADD to reach brooding onset than all of the other sites (Kahururu: x~ = 494 ADD, adj P < 0.001; Pakoka: x~ = 533 ADD, adj P < 0.001; Mangapiko: x~ = 480 ADD, adj P = 0.04; Table 2).

Significant differences in median ADD for E. aucklandica and E. menziesii to reach peak brooding (Stage 4) were found across all sites (Ohautira: P = 0.018, Kahururu: P = 0.01, Pakoka: P < 0.001, Mangapiko: P = 0.002). In E. aucklandica, median ADD for peak brooding varied significantly among sites (K–W: n = 162, H = 26.60, P < 0.001), with significant differences between Ohautira (lower ADD at peak brooding) and all other sites (Kahururu: P = 0.002, Pakoka: P < 0.001, Mangapiko: P < 0.001) (Table S3). Median ADD for peak brooding in E. menziesii also varied significantly among sites (K–W: n = 184, H = 31.95, P < 0.001), with significant differences occurring only between Mangapiko (higher ADD at peak brooding) and each of the other sites (Ohautira: adj P < 0.001; Kahururu: adj P < 0.001; Pakoka: adj P = 0.01; Table 2). The ADD required to complete the brooding period ranged from 1411 (Ohautira) to 1548 (Mangapiko) for E. aucklandica, and from 851 (Kahururu) to 1107 (Mangapiko) for E. menziesii (Table 2).

Beta regressions used to determine the strength of relationships between water temperature expressed as ADD and brooding onset and brooding peak in E. menziesii and E. aucklandica showed variable results. Significant relationships for brooding onset (Stage 1) were stronger in E. aucklandica (β = 0.010, s.e. = 0.004, P = 0.001) than E. menziesii (β = 0.005, s.e. = 0.002, P = 0.047), with marginal effect sizes averaging 0.2 percentage points in E. aucklandica and 0.1 percentage point in E. menziesii for every increase in ADD (Table S4 of the Supplementary material). Peak maturity (Stage 4) was negatively related to ADD in E. aucklandica (β = −0.001, s.e. = 0.001, P = 0.003), but there was no significant relationship for E. menziesii (β =−0.001, s.e. = 0.001, P = 0.2; Fig. 7). Here, average marginal effect sizes were −0.03 percentage points for every increase in ADD for both species (Table S4). Percentage variance explained by ADD for significant relationships ranged from 28 to 47% (Fig. 7).

Fig. 7.

Relationship (a, b) between ‘brooding onset’ proportion (glochidia Stage 1) and (c, d) between ‘peak brooding’ proportion (glochidia Stage 4) and number of accumulated degree-days in Echyridella aucklandica and E. menziesii. Solid lines show beta regression relationships that are statistically significant at P < 0.01, whereas the dashed line indicates P < 0.05. Pseudo r² values calculated using beta regression are given for each model.

Unionid demographics

All sites were dominated by older individuals, with no small mussels found below 40 mm for E. aucklandica or below 23 mm for E. menziesii. The presence of geriatric, senescing populations dominated by large individuals, as found in all study streams, has been observed within waterways throughout New Zealand (e.g. James 1985; Roper and Hickey 1994; Catlin et al. 2018; Melchior et al. 2023a) and abroad (Vaughn and Spooner 2004; Geist et al. 2006; Hastie 2011; Stöckl et al. 2014; Österling 2015), and provides evidence of recruitment constraints that could lead to ‘extinction debt’ (i.e. delayed extinction of species owing to past events or disturbances; Tilman et al. 1994; Vaughn 2012). Accordingly, the longevity of mussels (e.g. >50 years for E. menziesii, see Grimmond 1968; unknown for E. aucklandica), coupled with the scarcity of smaller mussels in these Waikato streams, could indicate that both species may not be recruiting sufficient numbers of juveniles to sustain viable populations over the long term. However, the mechanisms that affect unionid recruitment are complex and further studies are required to understand recruitment dynamics in both species, including host fish specificity (Melchior et al. 2021, 2023b).

In this study, the smallest female mussels collected in both species were found to be gravid. Thus, the minimum sizes for brooding maturity (23 mm in E. menziesii and 40 mm in E. aucklandica) inferred here are considered conservative estimates because the lack of juveniles found in the field meant that smaller females were not assessed. Nevertheless, the minimum sizes reported in this study fall within the normal range for age at maturity of unionids elsewhere in New Zealand and globally. For example, histological analyses of E. menziesii in Lake Taupo found the smallest sexually mature female to be 37 mm long, estimated to be ~4 years old (S. J. Clearwater, D. S. Roper, C. W. Hickey, M. L. Martin and E. K. Williams, unpubl. data). Small size at sexual maturity has also been reported in females of an Australian Hyriidae species Velesunio angasi, which attained maturity at 1.5 years of age (40 mm; cf. in 1 year or 30 mm length in males), whereas Hyridella species have been reported to mature in 2–3 years (40 mm) and Cucumerunio novaehollandiae at 3–4 years (60 mm) (Jones et al. 1986; Byrne 1998).

Interspecific variations in brooding phenology

Echyridella aucklandica began brooding earlier and remained gravid for longer than did E. menziesii in this study, reaching peak brooding (and thus glochidia release) in late austral spring to summer. Despite differing brooding onsets between species, both E. aucklandica and E. menziesii displayed synchrony in the peak brooding period, such that mature glochidia of both species were ready for release across all sites throughout the warmer part of the year in November and December, although timing appears to vary between regions, at least for E. menziesii (Percival 1931; Hine 1978), and waterbody types (e.g. stream v. lake; S. J. Clearwater, D. S. Roper, C. W. Hickey, M. L. Martin and E. K. Williams, unpubl. data). This general brooding seasonality is consistent with other hyriids found in temperate coastal Australia (New South Wales), including Hyridella australis, H. depressa and H. drapeta, which coexist in the Macleay River, and in the sympatric Velesunio ambiguous and A. jacksoni found in the Murray River (Jones et al. 1986; Walker 2017). As with E. menziesii and E. aucklandica, brooding in the above-mentioned hyriids was initiated in winter and glochidia were released in spring to summer (extending into autumn in the case of H. australis; Jones et al. 1986).

The extended brooding durations of both mussel species in the present study (although shorter for E. menziesii than E. aucklandica) are generally consistent with ‘bradyticty’ or a long-term brooding strategy, rather than ‘tachyticty’ or a short-term brooding strategy as recognised and described in many North American freshwater mussels (Sterki 1895, 1898; Ortmann 1911; Graf and Foighil 2000; Price and Eads 2011). In tachytictic mussels, fertilisation occurs in spring, with larvae brooded until they are fully developed and released within weeks to a few months over summer (Graf and Foighil 2000). Contrastingly, in bradityctic mussels, fertilisation usually starts in autumn, with brooding continuing over the winter, followed by the release of mature glochidia in spring (Graf and Foighil 2000). The important distinction between these patterns is that bradytictic mussels complete fertilisation and ontogenesis earlier and continue to brood long after glochidia are mature and viable (Kat 1984). Bradyticty, known only from temperate lineages of freshwater mussels, is hypothesised to be advantageous because earlier ontogeny permits earlier infestation and metamorphosis on hosts in spring, allowing more time for growth in juveniles before temperature reductions in winter (Graf 1997; Graf and Foighil 2000).

Although the bradyticty and tachyticty model is used widely in the literature to broadly classify unionid brooding strategies, some researchers have pointed out that the tachytictic or bradytictic classification is too simplistic and needs to be re-evaluated because brooding periods among unionid species can be highly variable (Kondo 1987; Watters and O’Dee 2000; Haag 2012). In support of this, the brooding durations exhibited in both E. aucklandica and E. menziesii showed interspecific variations across the study sites, a common feature found among coexisting species within the Hyriidae in Australia and South America (Jones et al. 1986; Callil et al. 2012; Walker 2017). Earlier brooding onsets and longer brooding durations in E. aucklandica than in E. menziesii were found to lead to earlier onset of glochidia maturation and longer duration of mature glochidia present in the marsupia of E. aucklandica. Although brooding peaks in both species overlapped, early brooding onset elsewhere is thought to facilitate the release of mature glochidia over a longer period (Haag and Warren 2000), potentially increasing the chance of successful host fish attachment for E. aucklandica

Varying developmental stages were present within the marsupia of both E. menziesii and E. aucklandica throughout their respective brooding seasons. Mature glochidia were present for up to 5 months in E. menziesii and 7 months in E. aucklandica, and the number of empty demibranchs was observed to decrease toward the end of the reproductive season, indicating that glochidia occurred over an extended period. This finding is consistent with other Hyriidae, including the Australian Alathyria jacksoni and V. ambiguous, which release their larvae throughout spring and summer (Jones et al. 1986; Walker 2017). However, it contrasts with findings for the northern hemisphere Margaritifera margaritifera by Hastie and Young (2003) and M. auriculata by Soler et al. (2019), who reported releases to occur as sudden synchronised events triggered by environmental cues such as water temperature or river levels (something we did not evaluate).

Longer glochidia release durations in unionids, as observed in E. aucklandica in Waikato streams, may occur for numerous reasons. First, they may reduce the risk of temporal mismatches with host abundances (Cushing 1990; Hochwald 1997; see Melchior et al. 2023b). Second, asynchronous, brooding (and thus glochidia release), found in both E. aucklandica and E. menziesii, may make both species less vulnerable to environmental disruptions, such as seasonal pulsed flow events.

Thermal cues associated with brooding

Our investigation indicated that thermal cues are important in regulating the timing of brooding onset, brooding duration, and peak brooding in both E. aucklandica and E. menziesii. Water temperature is known to influence the timing of phenological events by effects of thermal summation (i.e. ADD) (Davenport and Warmuth 1965; Young and Williams 1984; Hastie and Young 2003; Galbraith and Vaughn 2009; Dudding et al. 2019); however, minimum threshold temperature (Trübsbach 1998; Watters and O’Dee 2000; Blažek and Gelnar 2006; Schneider et al. 2018), and thermal shock (Jungbluth and Lehmann 1976; Hastie and Young 2003) are also processes that may be involved in different phenological phases (Jungbluth and Lehmann 1976; Hastie and Young 2003) and highlight the complexities of resolving thermal influences on reproductive phenology.

Generally, the ADD required to reach each of the phenological phases varied between species, with E. menziesii requiring broadly more ADD to reach brooding onset but less ADD for peak brooding and brooding duration than did E. aucklandica. Reasonably narrow ranges across sites in ADD required to reach brooding onset, particularly in E. aucklandica, coupled with a strong relationship between ADD and brooding onset in both species (although the strength of the relationship was weaker in E. menziesii), suggest that ADD may have some importance in regulating the timing of brooding onset in both species. Echyridella aucklandica brooding began in early to mid-winter when water temperatures were lowest among sites, whereas E. menziesii brooding onset occurred simultaneously across all sites in early spring when stream temperatures were increasing, suggesting an interaction with temperature thresholds as a secondary process initiating E. menziesii brooding. The negative relationship between peak brooding and ADD in E. aucklandica highlighted a gradual decline in the proportion of brooding unionids carrying mature glochidia with increasing ADD, probably indicating the onset of gradual larval release.

The proportion of brooding females in both species among all streams increased gradually over time, reaching peak brooding in November and December, before gradually declining until February and March. This synchronous pattern suggests that thermal summation (ADD) played a key role in determining brooding duration more so than a minimum threshold, which would have triggered sudden and simultaneous brooding or release events. The wide ADD ranges within species and between sites, particularly for peak brooding and brooding duration, along with large amounts of unexplained variation seen in the relationships of peak brooding and brooding onset with ADD, may be attributed to phenotypic plasticity in response to local environmental conditions (Bauer 1992; Johnson and Brown 2000; Hochwald 2001). For example, fish presence has been suggested to influence the timing of glochidia release, particularly in bradytictic brooding mussels, with some studies suggesting that host recognition mechanisms mediated by physical and chemical interactions may be important for the stimulation of glochidia release (Haag and Warren 2003; Barnhart et al. 2008; Schneider et al. 2018).

Melchior et al. (2023b) showed that E. aucklandica appears to use only Retropinna retropinna as a host fish, whereas E. menziesii is a host generalist, transforming on a range of native and introduced fish species, but not on R. retropinna. Timing glochidia release to coincide with host presence may be particularly important in New Zealand unionid species, where a large proportion of the known host fish species pool is diadromous. However, large flood events, which are usually associated with fish migrations, may override the impact of thermal cues, indicating that extended availability of glochidia may be an adaptive advantage where flooding is unpredictable.

As freshwater mussels are predominantly sedentary, with limited dispersal abilities as adults, they are especially vulnerable to the effects of a warming climate (Archambault et al. 2018). We have shown that some aspects of the reproductive phenology of southern hemisphere freshwater mussels appear sensitive to environmental cues such as water temperature. Specifically, shifts in temperature may potentially disrupt or shift reproductive timing in these species, with the potential for a disconnect between glochidia maturation and release and host fish abundances (Forrest and Miller-Rushing 2010; Schneider et al. 2018). Additionally, sudden fluctuations in temperature from a thermal regime to which species have adapted may impair mussel recruitment through abortions of immature and unviable larvae in early brooding females (Van Vreede et al. 1999; Schneider et al. 2018; see the ‘Context’, ‘Methods’ and ‘Results’ sections of the Supplementary material). It is therefore imperative to understand the potential impacts of climate change on both population- and community-level patterns in the reproductive phenology of unionid mussels and their reproductive synchrony with host fish.