This reflects the fact that total temporal variance of ES is generally higher at higher richness sites. All composition-loss terms were positive and all composition-gain terms negative, such that in all cases, composition terms partially cancelled their corresponding richness terms. This means that both the lost and gained species tended to have below-average contributions to total temporal variance of ES. Specifically, the positive composition-loss indicates that observed reductions in species richness resulted in less reduction in total temporal variance than would be the case if species losses had been random with respect to total temporal variance. Similarly, the negative composition-gain indicates that observed increases in species richness resulted in less increase in total temporal variance than would be the case if species gains had been random. In the watermelon system, composition-loss cancelled 86% of richness-loss and composition-gain cancelled 78% of richness-gain; in the blueberry system, these same value were 23% and 76%. Results were similar when total temporal variance was partitioned into variances and summed covariances . In both systems, between-site differences in abundance fluctuations of species present at both sites were more important than species loss and gain to between-site differences in total temporal variance of ES,nft hydroponic although only slightly so in the blueberry system.

Specifically, in watermelon only 9% of between-site differences in total temporal variance of ES was attributable to species loss and gain, while the remainder was attributable to between-site differences in the abundance fluctuations of the species present at both sites. For blueberry, 42% of the total temporal variance was attributable to species loss and gain, with abundance fluctuations of common species accounting for the remainder. Results were similar when total temporal variance was partitioned into variances and summed covariances, and in the senstivity analysis in which . We found that abundance fluctuations of common species, rather than species richness per se, was the stronger driver of total temporal variance of an ES. This result was especially strong in the watermelon system, and less-pronounced in the blueberry system. Thus, our results contribute to a growing collection of studies that emphasize the importance of abundant, broadly-distributed species to ecosystem services . Abundance fluctuations accounted for 91% and 58% of the between-site differences in total temporal variance in pollination for watermelon and blueberry, respectively, with the remaining 9% and 42% accounted for by losses and gains of species. There are two reasons for this result. First, the skewed species abundance distribution results in few highly abundant species capable of large abundance fluctuations and many rare pollinator species, whose abundance fluctuations are constrained by their rarity. For example, singletons and doubletons combined account for 28% and 45% of the species in the watermelon and blueberry data sets, respectively. Second, rare species were more likely to be lost and gained between sites, accounting for most changes in richness between sites.

As a corollary, the highly abundant species with the greatest impacts on the variance in pollination were present at all or most sites, and the variance they contributed to ES was generally not attributable to species loss or gain between sites. Although it is possible that richness could determine ES if dominant species were more likely to be found at higher-richness sites, we found the dominant species were present regardless of species richness. This contrasts with randomly assembled biodiversity-ecosystem function experiments in which richness increases the chances of including a high-ES species that may then become dominant in mixtures . Thus, for the ES we studied, richness and temporal variation of ES are decoupled from each other, because the rare species accounting for the richness changes contribute little to the variance. While observational studies in un-manipulated systems could examine the shape of the richness variance relationship without considering composition, the result would be misleading because not only composition, but also evenness and abundance co-vary with richness but are not captured by the x-axis . The Price equation approach resolves this issue by defining effects of richness changes as those that are random with respect to species’ contributions to total temporal variance, such that the Price equation richness terms are not confounded with composition . The composition terms capture only those effects of species loss and gain that are nonrandom with respect to species’ contributions to total temporal variance, and so depend on the identity of lost and gained species. Accordingly, the Price equation does not attempt to replicate the approach used in biodiversity-ecosystem function studies . Rather, it provides a different perspective on the question that originally motivated much biodiversity-ecosystem function research: how will ongoing species losses affect ES? In our study, the Price equation allowed us to separate richness and composition effects, showing that because of the non-random identity of lost species – namely, they tended to be rare species with little impact on temporal variance – reductions in richness did not greatly affect temporal variance of ES.

Synchrony or asynchrony of species responses to environmental changes potentially has a major impact on the stability of ES, with asynchronous fluctuations stabilizing ES over time and synchronous fluctuations increasing the temporal variance of ES. We found that native bee species’ abundances exhibit positively-correlated fluctuations . While we might expect competing species to show compensatory dynamics , competition does not appear to be a major determinant of population dynamics in bees , and therefore the lack of negative covariances is unsurprising. Because of positive correlations between species’ variances and summed covariances, our analyses produced very similar results regardless of whether we analyzed variances or summed covariances by themselves, or total temporal variance . In other words, changes in richness and composition across sites had little effect on aggregate covariance, because the species accounting for these changes were rare and had low summed covariances. In contrast, within each system, species with high, positive covariances were present at nearly all sites, regardless of species richness, and explained most between-site differences in aggregate covariance. Some mechanisms through which richness or composition could reduce the variance of ES are excluded by our methods, and would need to be investigated with more detailed field studies. For example, because our estimates of per-visit pollen deposition are based on one individual bee visiting a virgin flower, our methods may not capture how multiple bee species sequentially visiting a plant may lead to complementarity in pollen deposition,nft system either in space or in time . Likewise, for each pollinator taxon we apply a pooled value of per-visit pollen deposition rather than allowing this to change among sites. Potential functional compensation among species is thus merged into the ABUN term in our approach. Similarly, we exclude any factors that disrupt the relationship between pollen deposition and fruit yield; however, a recent analysis of hundreds of crop fields found that pollen deposition is generally a good proxy for fruit set . The method requires multiple years of information on both species abundance and species efficiency across spatial replicates. In some cases, it may not be logistically feasible to obtain efficiency estimates for every individual in the study. Here, we estimated efficiency using single-visit experiments and others wishing to apply this method will need to develop their own system-specific method of estimating efficiency. In some cases , this could involve destructive sampling or allometric equations that estimate biomass from height and other characteristics. In conclusion, we developed and used a new analytical method to show that abundant species, not species richness, is the stronger driver of total temporal variance of ES in two pollination systems. This new method is the first to partition between-site differences in the total temporal variance of ES into terms attributed to richness, composition, and abundance. Here, abundance had a strong effect because—as is true for nearly all ecological communities—our study systems contained few abundant and many rare species, and because between-site differences in richness were due to rare species that did not greatly affect total temporal variance of ES. The Price approach as developed here is a broadly-applicable framework, which can be used to analyze the temporal variance of any ES that is expressible as a sum of species contributions ; for example, above ground carbon storage in forests or fisheries yields.

Given its generality, we believe this method can help better resolve pressing questions about the relative importance of richness and abundance for the stability of ecosystem services. Human-altered landscapes are expanding globally and are often associated with declining natural habitat, non-native species, fragmentation, and transformations in structure, inputs, climate, and connectivity. These changes collectively have resulted in shifts in both spatial distributions and species diversity across many taxa including birds, mammals, reptiles, amphibians, invertebrates, and plants. One common driver of global change is urbanization, which in the extreme is associated with a reduction in biodiversity compared to habitats in their more natural state. However, in moderately urbanized areas, the effects of urban impacts on species distribution and diversity can vary greatly and depends on region, type of change, and taxonomic group, among other factors. Documenting the effects of urbanization compared to natural communities has proven problematic, making predictions of community change associated with urbanization difficult. Human-altered landscapes are often associated with many non-native species which add to species diversity but also can obscure changes in community dynamics. Thus, to assess accurately the complex impacts of land use change on ecological communities, one must look beyond species richness to investigate ecological processes themselves. Ecological processes are the links between organisms in a functioning ecosystem, and are critical in understanding how altered biodiversity can lead to changes in ecosystem functioning. Global environmental change has been found to have a wide variety of impacts on ecological processes in different systems. Pollinator-plant relationships in particular are found to be particularly vulnerable to land use change, resulting in decreases in interaction strength and frequency. Pollination services are crucial ecosystem processes in natural systems, but also in agricultural and urban areas. Bees provide the majority of animal-mediated pollination services on which it is estimated 87.5% of flowering plants depend. The value of pollination in agriculture is estimated at $200 billion worldwide, largely due to many foods that are essential for food security and a healthy human diet, including numerous fruits, vegetables, and nuts that require bee pollination. As urban areas expand, there has been increasing interest in urban agriculture to ensure food security and access to healthy foods for growing populations, and these systems also depend on pollination. For example, Kollin estimated that the economic value of urban fruit trees to be worth $10 million annually in San Jose, California. Despite the important role of pollinators and concerns about bee declines, there remain many uncertainties regarding the impact of land use change on pollinators. Urbanization has resulted in more interfaces with both natural and agricultural landscapes, creating new transitional zones of peri-urbanization. While there has been extensive pollinator research in agricultural and natural systems, less attention has focused on pollination in neighboring urban areas and how the changing landscape has impacted pollination. In addition, very few studies of urban areas have looked beyond changes in bee diversity to understand explicitly the effect of urbanization on pollinator-plant interactions. Here, we investigate the effect of land use change on pollinator plant ecosystem processes. We make use of a ‘‘natural experimental design’’ in which urban, agricultural, and natural areas intersect. Bees visit flowers for both pollen and nectar resources, and floral visitation is a commonly used as an index of pollination services. However, depending on the flower, certain bee groups are much more effective pollinators than others. Thus, while visitation is important, it alone does not definitively indicate whether pollination services were received by the plant. When pollen is limited by other factors, consequences for plant fitness can include failure to set seed, production of smaller fruits, and even complete lack of reproduction. By looking at rates of bee visitation and comparing this with other measures of plant fitness, such as seed set, we can develop a more complete understanding of how shifts in bee distributions between areas that differ in land use are impacting pollination services. To study the impact of changing land use on pollinator-plant interactions, we focus on bee pollination of a widespread plant, yellow starthistle , a common weed found in natural, agricultural, and urban habitats.