Pollinator competition and the structure of ﬂoral resources

3 The mutualism between plants and pollinators is built upon the trophic ecology of flowers and florivores. 4 Yet the ecology of flowers-as-food is left implicit in most studies of plant-pollinator ecology, and it has been 5 largely neglected in mainstream trophic ecology. This deficit is especially evident in an emerging issue of basic 6 and applied significance: competition between pollinators for floral resources. In this synthesis, we start by 7 exploring the notion of floral resource limitation upon which most studies concerning competition between 8 pollinators are tacitly predicated. Both theoretical and empirical lines of evidence indicate that floral resource 9 limitation must be understood as a complex ecological contingency; the question is not simply whether but 10 when , where , and in what regions of floral trait space resources are limiting. Based on this premise, we propose 11 a framework for understanding floral resource availability in terms of temporal, spatial, and functional 12 structure. While this framework is conceptually intuitive, it is empirically and analytically demanding. We 13 review existing methods for measuring and summarizing the multi-dimensional structure of floral resources, 14 highlight their strengths and weaknesses, and identify opportunities for future methods development. We

abundant in a landscape, such as a maple forest or a rapeseed field at full bloom, but their abundance is not 140 stable like that of foliage; the same forest and field could be almost flowerless a few weeks after peak bloom 141 (Requier et al. 2015). The notion of intactness is also more nuanced for flowers than for foliage, since the 142 depletion of nectar and pollen is not visually apparent. A landscape dense with flowers could nevertheless 143 be depauperate in nectar and pollen if the flowers have been heavily exploited (Heinrich 1976) or if their 144 productivity has been stunted by drought (Waser and Price 2016, Phillips et al. 2018). As for robustness to 145 weather events, flowers are presumably more sensitive than foliage (e.g. Papadopoulou et al. 2018), though 146 the topic is not well-studied. Thus, the premises of the green world hypothesis, when extended to flowers 147 and florivores, appear to be at best contingencies rather than givens. 148 The classical objections to green world hypothesis are, however, similarly problematic in the context of  159 Kouřimská and Adámková 2016), suggesting that pollen-feeders have more in common with carnivores than 160 with folivores when it comes to nitrogen nutrition. 161 Considered over evolutionary time scales, it has been hypothesized that there should exist a positive feedback 162 loop wherein food scarcity for pollinators entails visitation saturation for plants, resulting in selection pressure 163 toward lower investment in nectar production, and thus more extreme nectar scarcity (Ratnieks and Balfour 164 2021). Indeed, precisely this evolutionary effect has been reported in the alpine lotus (Saussurea nigrescens) 165 in response to high densities of managed honey bees, and the evolutionary process was rapid enough to be 166 detected over the course of just three decades (Mu et al. 2014). Importantly, though, the logic of positive 167 feedback works in both directions; when nectar is non-limiting, plants can be expected to compete for limited 168 pollinator visitation, generating selection pressure toward increased nectar production (Ratnieks and Balfour 169 2021) and ultimately "sweet world" conditions. Indeed, the fact that mechanisms for nectar resorption are  Thus, nectar-and pollen-feeding, while clearly special cases of herbivory, do not map neatly onto the classical 173 debate concerning resource limitation in herbivores, and this uncertainty is exacerbated by the expectation 174 of destabilizing evolutionary feedback between the production and consumption of floral resources. Given this equivocity of theory, it is perhaps not surprising that empirical studies have reported both apparently 176 limiting and apparently non-limiting conditions, often alternating through diel or seasonal time within a 177 single locality (Hocking 1968, Mosquin 1971, Heinrich 1976, Roubik and Buchmann 1984, Bowers 1986, evidence indicates that floral resource limitation is the most common constraint on pollinator populations.  It can be said that an organism is "food-limited" if an increase in food availability -either due to increased 192 food supply or the release of food from competition -would increase the organism's fitness (i.e. its repro-193 ductive success). By extension, a food-limited population is one whose rate of growth would increase with 194 increased food availability. Importantly, this definition does not require the exhaustion of food resources 195 or the actual starvation of individuals. Since time spent foraging entails energy expenditure, risk of pre-196 dation, and the delay of other vital activities (e.g. mating, oviposition, nest construction and defense), it 197 can be expected that fitness will generally increase with the temporal rate of food acquisition -indeed, 198 this expectation is a key premise of optimal foraging theory (Fretwell and Lucas 1969) and its application 199 to insect pollinators (Goulson 1999). Food acquisition rate will, in turn, increase with resource availability Figure 1: Spatial, temporal, and functional dimensions of floral resource structure. For any spatial unit (A), floral resource availability varies through time at both seasonal (B) and diel scales (C). These patterns, in turn, vary through space, and aggregate floral resource availability at any given time and place is distributed across the functional variation of the floral community (e.g. flower shape, corolla depth, color), which can be represented in discrete form as functional compartments (e.g. zygomorphic flowers, deep flowers, violet flowers).

Production, consumption, standing crop, and depletion
Before elaborating the dimensions of time, space, and functional traits in which floral resources are struc-214 tured, it is important to distinguish four interlocking senses in which floral resource conditions can be 215 understood: production, consumption, standing crop, and depletion. Floral resource production, the rate   In the sections below, our primary interest is in the structure of the standing crop and the depletion rate of 235 floral resources, since these concepts bear the most direct relation to pollinator coexistence, though we will 236 also touch on the underlying processes of floral resource production and consumption. We emphasize that 237 each of these patterns and processes can be mapped onto the dimensions of time, space, and functional traits 238 presented in our conceptual framework. We will revisit the distinction between production, consumption,    the species level, the timing of peak pollen presentation varied broadly, ranging from early morning (before 274 9:00) to late afternoon (16:00), and in rare cases even during the night. Aggregated across plant species, 275 though, community-level pollen presentation was generally unimodal and peaked between 8:00 and 11:00, 276 with seasonal variation. Notably, Percival (1955) also recorded concomitant pollen foraging by honey bees, 277 and she found that it tended to be shifted later in the day by roughly two hours relative to the start and peak Behind seasonal and diel patterns of floral resources there also exist processes that influence floral resources 288 over supra-annual time scales. We will revisit this topic in Section 5.3.  Goulson 1999). So, from the perspective of a given species at a given time and place, spatial structure in 300 floral resource production is, in a sense, behaviorally averaged out into a more or less uniform standing crop.

301
This effect is perhaps most intuitive in the colony-level foraging behavior of eusocial species like honey bees 302 and bumble bees, which have indeed been shown to allocate foraging effort in a manner that approximates 303 the predictions of OFT ( Bartholdi et al. 1993, Dreisig 1995. In principle, though, the theory applies also 304 to solitary species at the population level.

305
OFT depends on idealizing assumptions that never obtain perfectly in real systems, including that foragers 306 have perfect knowledge of their environment and that there exist no constraints (e.g. interference compe- stably. Moreover, while we have assumed the approximation of OFT, with its tendency to negate spatial 318 heterogeneity in floral resources, it is important to remember that the spatial structure of floral resources 319 interacts with the temporal dynamics described above, which may prevent the equilibrium conditions of 320 OFT from being realized. When temporal dynamics are considered in conjunction with spatial heterogene-321 ity, the process of patch discovery can be decisive in determining foraging efficiency (Visscher and Seeley 322 1982, Schürch and Grüter 2014), invoking the classical concept of "fugitive species" and the potential for a 323 stabilizing tradeoff between colonization (in this case, patch discovery) and dominance (Hutchinson 1951, 324 Hanski 1995).

325
Finally, it is important to note that our discussion thus far has assumed central place foraging. For non-  interwoven with that of the pollinators that feed on them, raises an important distinction that has thus far 361 remained latent in our discussion of competition, namely the crucial difference between inter-and intra-362 specific competition. It is this distinction that defines the relationship between competition and coexistence, 363 since the fundamental condition for coexistence is not the absence of competition but rather that competition 364 within species exceed competition between species, such that each species limits its own population density 365 more than it limits that of other species (Hanski 1995). Whether this condition is met depends on the degree   Figure 2: A three-domain research agenda for the study of floral resource structure. Empirical and analytical techniques are needed for measuring floral resource structure (A) and deriving meaningful summary metrics, such as temporal maxima and minima (B), slopes over discrete time frames (C), dimensionality reduction via ordination (D), and full-dimensional analysis using hypervolumes (E). When appropriately measured and summarized, floral resource structure can be studied inferentially as a modulator of competition and coexistence (F) and ultimately a driver of pollinator community composition (G). Conversely, floral resource structure is also an effect of pollinator community composition, together with a suite of exogenous drivers (H). 389 Measuring floral resources at spatial and temporal scales relevant to pollinator foraging is a long-standing  Portions of the curve above zero (dotted line) indicate weight gain while portions below zero indicate weight loss. This study system (Philadelphia, PA) exhibits the classic pattern of summer dearth (mid-August) that has been described in many temperate systems. Two strong pulses of floral resource abundance are evident in the spring, and a brief late pulse occurs after the summer dearth. Alternatively, similar inferences might be achieved via a floral indicator species. As discussed earlier, optimal 438 foraging theory (OFT) predicts that foragers will distribute themselves so as to equalize individual rate of 439 reward across resource patches, and there is some evidence that flower-visiting insects approximate this

583
Uniting pattern and process is the crux of application. Clarifying the causal relationships between floral 584 resource structure, pollinator coexistence, and plant-pollinator interactions -as well as the sensitivity of 585 each to exogenous drivers -will provide a coherent basis for addressing contentious management issues, 586 such as the compatibility of apiculture with wild pollinator conservation and the appropriate use of floral 587 enhancements in agri-environment schemes.

588
With regard to the specific issue of potential competition between managed honey bees and wild pollinators, 589 one very practical implication of our discussion is that it would be prudent to evaluate floral resource standing 590 crop and depletion rate (see Section 5.1), at relevant times of year, when considering the introduction of 591 honey bees to a given locality. This kind of site assessment protocol could complement other approaches 592 to apicultural regulation (e.g. Henry and Rodet 2020), and the information gained would be as useful to 593 beekeepers as to conservationists, since neither party benefits from the addition of colonies to an already