Bee phenology is predicted by climatic variation and functional traits

Abstract

Climate change is shifting the environmental cues that determine the phenology of interacting species. Plant-pollinator systems may be susceptible to temporal mismatch if bees and flowering plants differ in their phenological responses to warming temperatures. While the cues that trigger flowering are well-understood, little is known about what determines bee phenology. Using generalised additive models, we analyzed time-series data representing 67 bee species collected over 9 years in the Colorado Rocky Mountains to perform the first community-wide quantification of the drivers of bee phenology. Bee emergence was sensitive to climatic variation, advancing with earlier snowmelt timing, whereas later phenophases were best explained by functional traits including overwintering stage and nest location. Comparison of these findings to a long-term flower study showed that bee phenology is less sensitive than flower phenology to climatic variation, indicating potential for reduced synchrony of flowers and pollinators under climate change.

Experimental Test of the Combined Effects of Water Availability and Flowering Time on Pollinator Visitation and Seed Set

“…Climate change can also influence abundance or phenology of the insect pollinator populations themselves in this locality, in part due to indirect effects of floral resource phenology (Ogilvie et al, 2017), and climate-induced changes in insect populations were not included in our manipulations. Comparisons of bee phenology across time in this locality suggest that emergence of solitary bees is advancing by 4.9 days per decade (Stemkovski et al, 2020), which may be faster than that of M. ciliata. Syrphid fly emergence responds to date of snowmelt, making earlier emergence likely in the future even though it has not yet been demonstrated (Iler et al, 2013).…”

Geographic variation in the robustness of pollination networks is mediated by modularity

“…More importantly, we revealed the effect of insularity on the environment–network relationship for the first time: PS showed opposite effects on modularity between island and mainland networks. An explanation might be as follows: species vary in their sensitivity to seasonal change (Stemkovski et al., 2020), which means some species have a long phenophase and they usually are generalists, such as bumblebees, while others do not (Bascompte & Olesen, 2015; Olesen et al., 2008). So the opposite effect of PS on modularity may depend on the proportion of long‐phenophase generalists in the communities: in communities with few long‐phenophase generalists, most species can only interact with a few co‐occurring partners in a short time period and they form several phenological units, thus seasonality would promote module partitioning by minimizing the time overlap of the phenological units (Bascompte & Olesen, 2015; Martín González et al., 2012); but in communities with many long‐phenophase generalists, seasonality would make the generalists become connectors among modules because the generalists co‐occur with several phenological units across time (Bascompte & Olesen, 2015; Martín González et al., 2012).…”

Estimating abundance and phenology from transect count data with GLMs

“…However, because our proposed approach fits a Gaussian curve, from estimated µ and σ it is trivial to calculate any characteristics of a Gaussian curve, including (a) any arbitrary quantiles (e.g. the 0.05 and 0.95 quantiles used in Stemkovski et al 2020), (b) the height of the curve (e.g. maximum number of flowers, Miller‐Rushing and Inouye 2009) or (c) the ‘observable flight season' (days when the curve exceeds 1, a metric reflecting the period of likely human detection that parallels first and last observation dates) (Bonoan et al 2021).…”

“…One of the common forms of sampling for insect populations is systematic repeated surveys throughout an activity period, such as ‘Pollard' transect walks (Pollard 1977), bee bowls (Stemkovski et al 2020) or trap nests (Forrest and Thomson 2010). Historically, the main goal of these surveys was simply to estimate yearly abundance (Zonneveld 1991, Pollard and Yates 1993, Schultz and Hammond 2003).…”

“…Perhaps as a consequence of being developed with such rich data, current methods generally require considerable data to work. This limitation holds both for the suite of models developed for flowering plants, the ‘Zonneveld model' – a mechanistic phenology curve commonly used to analyze insect counts (Zonneveld 1991, INCA 2002, Haddad et al 2008) –, and more generic approaches like generalized additive models (GAMs) (Rothery and Roy 2001, Hodgson et al 2011, Newson et al 2016, Stemkovski et al 2020).…”