Discussion
The timing of resource abundance relative to colony development profoundly impacted colony growth and reproduction of B. vosnesenskii . Colonies that received supplemental resources earlier in development achieved greater peak weights and also produced more queens – on average over 10 times more – than those colonies that received supplemental food later. Greater peak weights of early-pulse colonies were not driven by greater growth during a simulated resource pulse. Instead, early-pulse colonies had higher growth rates during the period that followed, in which they had access to ambient resources only. The increase in growth rate is consistent with expectations based on transient dynamics. Past studies have shown that higher food resources increase bumble bee body size (Sutcliffe & Plowright 1988; Persson & Smith 2011; Malfi et al. 2019), and that larger bumble bees bring back more resources (Goulson et al. 2002; Spaethe & Weidenmüller 2002; Klein et al. 2017; Kerr et al. 2019) which would lead to higher colony growth rates after the pulse ended. Our early-pulse colonies had higher foraging rates compared to late-pulse colonies during their off-pulse periods, even when controlling for colony size, suggesting there was a more effective foraging workforce during the post-pulse period in early-pulse colonies (Fig S4). It is also possible that elevated resources may have led to additional carry-over effects beyond the change in body size structure per se . For example, Malfi et al. (2019) found that elevated resources increase bumble bee worker longevity, as well as body size. Conversely, captive workers deprived of pollen died earlier than those allowed to feed ad libitum on pollen (Smeets & Duchateau 2003). Further investigation of these mechanisms would be an interesting avenue for future research.
In contrast to persistent pulse effects in early-pulse colonies, we did not see a corresponding negative carryover effect of the off-pulse period in the late-pulse colonies, in that their growth rates were not different from those of early-pulse colonies during the pulse itself. One hypothesis for similar responses to the pulse is that early-pulse and late-pulse colonies had the same distribution of worker sizes at the start of the pulse; in other words, all colonies had similar body-size distributions when placed in the field regardless of treatment, and late-pulse colonies might have retained this distribution during the off-pulse period. Another possibility is that, because supplemental resources were added by hand to colonies, size-based foraging ability did not strongly affect resource gain during the period of supplementation; in other words, worker size distribution did not matter. Foraging levels were equally low during pulses for early-pulse and late-pulse colonies (Fig S4). By supplementing colonies at the nest, we also likely affected other aspects of colony dynamics, e.g., mortality associated with foraging (Rodd et al. 1980; Cresswell 2017; Malfi et al. 2018). A third hypothesis is that, during the pulse, colonies were limited by factors other than resource return. For certain Hymenoptera in particular, fecundity is sometimes limited by the rate of egg maturation (Zhang et al. 2014; Yadav & Borges 2018), rather than resource return, particularly under high resource scenarios (Heimpel & Rosenheim 1998; Neff 2008).
Despite the strong carryover effect of resource pulses on off-pulse growth rate, the timing of resource pulses did not significantly affect the switch point (τ). Optimality models of bumble bee colonies (Macevicz & Oster 1976; Beekman et al. 1998) predict the timing of reproduction for bumble bee colonies should be earlier in high-resource environments. In these models, the key effect of resource abundance is on whether colonies are limited by queen egg production (presumably, high resource environments), which leads to earlier switches to reproduction, or limited by worker resource return (presumably, low resource environments), which leads to later switches to reproduction. Consistent with these models, Bowers (1986) observed that B. flavifrons colonies in high-resource meadows switched to reproduction sooner than those in low-resource meadows. It may be that bumble bee colonies respond differently to fluctuations in resource limitation (our experimental pulse treatments) than to chronically high or low resource conditions. It may also be that B. vosnesenskii life histories were resource limited (not egg limited) in both food resource treatments, and so timing of the switch from growth to reproduction was cued by end-of-season environmental conditions in both cases.
The availability of floral resources within landscapes varies throughout the season with pulses and dearth occurring at different times (Reader 1984; Liz & Ruiz-Herrera 2016). In our study region, the Central Valley of California, there is a steady decline of flowers from the onset of the growing season to the dry season in natural areas, with mass-flowering crops producing late season resource pulses in some agriculturally dominated areas (Williams et al. 2012). In a European agricultural system, late season forage from mass-flowering clover increased queen abundance, but early season mass-flowering did not elevate queen numbers (Rundlöf et al. 2014) or colony-level queen production in similar landscapes (Westphal et al. 2009). It is tempting to speculate that persistent effects of the early-season pulse in our system reflected dynamics of ambient resources, specifically elevated late season resources in some landscapes. Indeed, new queen production was highest in four colonies that both received early supplements and encountered a natural increase in late-season ambient forage. However, analyses of growth rates across all colonies suggest that ambient resources are not the primary driving factor behind carry-over effects of the early-season pulse (Supporting Information, Appendix 2 ).
More generally, our work emphasizes the need to understand the temporal distribution of available flowering resources and its influence on population dynamics of bees. Pollinator conservation efforts are often based on planting floral resources to achieve greater abundance of forage resources but must also recognize the importance of resource continuity and timing (Schellhorn et al. 2015; Scheper et al. 2015; Williams et al. 2015). Our study strongly indicates the importance of resource timing for bumble bee colony and population health and supports targeted efforts to boost resources during the early season in areas where they are lacking, such as the agricultural lands in our study system (Williams et al. 2012). At its simplest, our study indicates that the timing of resources affects both the phenology of colony growth and the ultimate reproductive output, so season-long estimates of floral resources are not an adequate metric of habitat quality.
In closing, our results emphasize the importance of the timing of resources for colony growth, reproduction, and phenology. Carry-over effects have been widely demonstrated in plant and animal populations, but only rarely linked to population dynamics (Beckerman et al.2002). For the particular case of transient dynamics, effects of changes in size structure have mostly been evaluated using models (McDonaldet al. 2016), not field experiments where populations are confronted both with large perturbations and ordinary environmental fluctuations. Numerous landscape studies of bees focus on the spatial context of land use and resource availability, whereas extremely few have investigated temporal dynamics within sites. Conservation planning for bumble bee populations will need to emphasize not only overall quality habitats but the temporal pattern of resources they contain. Our results also demonstrate the general importance of temporal variation for population dynamics, and the utility of bumble bee colony growth as a model system for understanding population dynamics in temporally varying environments.