Using standardized observations of floral visitation and seed set measurements of yellow starthistle, we test the hypotheses that increasing urbanization decreases 1) rates of bee visitation, 2) viable seed set, and 3) the efficiency of pollination . In addition to contributing to a better understanding of how change in landscape use, particularly urbanization, affects pollination-plant interactions, the study illustrates the importance of use of neighboring lands for pollination services.Our study system was located around Brentwood, in east Contra Costa County, California, where natural, agricultural, and urban areas intersect with each other within a 20620 km region . A county water district , regional park district , and California state park all fall within the region, leaving large areas of land protected from development. This protected land consists mainly of grasslands and oak woodlands, some portions of which are managed for grazing. East Contra Costa County has had a farming community presence since the late 19th century. The agricultural areas of Brentwood, Knightsen, and Byron mostly consist of orchards , corn, alfalfa, and tomatoes. A housing boom in the 1990s led to massive residential growth in the area. The city of Brentwood has grown from less than 2500 people in the 1970s to over 50,000 today , and nearby Antioch has now over 100,000 residents .We selected 12 sites dominated by yellow star thistle in a stratified design to span the different land use types . Yellow starthistle is a common weedy plant that forms homogenous flowering patches in grassy areas throughout this region. Many different bee taxa in a range of functional groups and size classes have been observed to visit yellow starthistle, in part because it flowers late in the season relative to other floral resources. Despite being considered a serious introduced weed, yellow starthistle is unusual as an invasive species in that it depends on animal pollinator visits in order to set seed. Within each site we selected a 50 m650 m plot such that each plot was at least 2 km away from all others,hydroponic gutter a distance larger than the maximum assumed typical bee foraging ranges.

Although certain bee species have been recorded foraging as far as 1400 m, most bees in this type of habitat have nesting and foraging habitat within a few hundred meters of each other. Within each plot we estimated number of flowering yellow starthistle blooms by randomly placing 10, 1 m61 m quadrats and counting the number of flowering blooms in each. We also measured the spatial area of yellow starthistle patches within each 50 m650 m plot to obtain an estimate of total flowering blooms within each plot. We categorized total blooms/plot on a log scale: ,103 , 103 –104 , and .104 . Using NOAA’s 2006 Pacific Coast Land Cover dataset , a 500 m buffer was created around each plot, and the number of pixels classified as agricultural, urban, natural, water, or bare land was extracted. These categories were obtained by lumping finer categories in NOAA’s classification scheme using the following definitions: Urban–‘‘High Intensity Developed’’, ‘‘Medium Intensity Developed’’, ‘‘Low Intensity Developed’’, and ‘‘Developed Open Space’’; Agricultural–‘‘Cultivated’’, ‘‘Pasture/ Hay’’; Natural– ‘‘Grassland’’, Deciduous Forest’’, ‘‘Mixed Forest’’, ‘‘Scrub/Shrub’’. Each plot was classified as a proportion of each of the 3 different land use categories, as well as for the category that was dominant. By this latter measure, of our 12 sites, 4 of each were classified as ‘‘urban’’, ‘‘agricultural’’, and ‘‘natural’’.We observed visits by all bee species to yellow starthistle at all sites 3 times for a 30 min period for a total of 90 min of total observation time per site within the same 2 wk period in August 2011. AM was defined as being between the hours of 9:30–11:30, Mid-Day as between 11:30–13:30, and PM as between 13:30–15:30. Also recorded at each observation period were approximate number of blooms, and wind and temperature simultaneously . Bees were not netted for later identification as we did not want to interfere with visitation to starthistle during this study. Instead, we used a modified protocol of citizen scientist observation surveys with 15 expected bee morphotypes that correspond to 30 possible genera known to occur in the region .

The observer slowly walked through the yellow starthistle patch, and upon reaching patch edge, returned on a path at least 3 m away from the previous, and recorded the morphotype classification of all bee visitors within 1.5 m on either side of the transect.Yellow starthistle has composite flowers, which are aggregations of anywhere from 20–80 florets. At each site, 12 yellow star thistle buds were randomly selected from different plants and covered with a mesh bag. Yellow starthistle blooming cycles have been described in detail in other publications. We selected buds at stage BU-4, when buds had no yellow petals exposed, but had well-developed straw-colored spines. When in full flowering, 10 bags were opened for a 4 hour period from 10 am to 2 pm, while 2 were kept closed as controls to verify that self-pollination was not occurring. At the opening and re-closing of the bags, the number of florets that had their stigmas extended were counted. Later, when flowers were fully mature , seed heads were collected, and later dissected in the lab. Viable and non-viable seeds in yellow starthistle seed heads are easily distinguishable based on color and shape. Because yellow starthistle requires pollination to produce viable seeds , non-viable seeds represent pollen limitation occurring during the 4-hour period that the flowers were exposed to pollinators. All seeds were counted to compare ratios of viable to non-viable seeds. Any seed predation was noted, and when possible, the seed predator was identified.All analyses were done in R 2.15.1 . Because each site had an AM, Mid-Day, and PM observation event, there were a total of 36 observation events, each with unique wind and temperature recordings, and visit observations of the 15 bee morphotypes. From these, we calculated the total number of bee visitors, total number of bee morphotypes, Shannon diversity of morphotypes, and morphotype evenness. Shannon diversity and evenness were calculated using the R package vega. The spatial auto-correlation of all bee visitor response variables was assessed by Mantel tests in R package ade4, using the average values for each time of day at each site. Spatial auto-correlation was not detected . To test for the effect of land use type on each of the response variables we used a generalized linear mixed model using the R package lme4. We designated land use type, bloom category of flowering patch, observation time period, wind, and temperature as fixed effects and site as a random effect.

Natural land use and AM observation time period were the model baselines for the categorical variables of land use type and observation time. Shannon diversity and evenness were fit with Gaussian distributions while all other variables were fit with Poisson distributions. In comparing the ratios of viable seeds to total seeds vs. the ratio of viable seeds to counted stigmas, we found that there was a strong correlation between these metrics. To look at the effect of land use type on seed-set, we therefore decided to utilize the ratio of viable seeds to total seeds in each seed head that did not experience seed predation, because of error in counting the number of stigmas . We then used a generalized linear mixed model fit with a Binomial distribution, with land use type as a fixed effect and site as a random effect. Finally,u planting gutter we tested for an effect of floral visitor observations on yellow starthistle seed set at each site. We averaged the number of visits from each morphotype across temporal observation events at the same site. Morphotypes that averaged at least one visit per 30 minute observation window were included as fixed effects in alinear mixed model fit with a binomial distribution, with site as a random effect and the ratio of viable to total seeds as the response variable. We also modeled the effects of total bee visitation, morphotype richness, and morphotype diversity on seed set ratios.Our results show that rates of bee visitation and seed set vary among urban, agricultural, and natural landscapes, demonstrating the importance of land use in the dynamics of plant-pollinator interactions. We suggest that these effects are at least in part explained by floral availability, a vital bee resource, which can be highly variable among different land use types. For example, in August there are few plants in flower besides yellow starthistle in the natural areas of Contra Costa County, California, whereas in urban and agricultural areas there are many exotic plants and supplementary inputs available . From pantrapping of bee specimens in the region , we know that total bee abundance is highest in the spring in natural areas. However, towards the end of the summer when yellow starthistle is in flower, there is little difference in collected bee abundance between human-altered landscapes and natural areas, and human-altered areas may even exhibit overall higher bee abundance. Our results of bee visitation to yellow starthistle support this pattern. Agricultural areas have large populations of managed honey bee colonies, so one would predict visitation to yellow starthistle by honey bees to be positively associated with surrounding agricultural land use. By contrast for native bees , the highest rates of visitation to yellow starthistle were in sites with more surrounding urban land use. Urban gardens have many exotic plants, often selected for aesthetic purposes, many of which are in flower later in the season than most California native plants. In addition, many of the plants in urban areas both directly and indirectly receive supplementary resources, particularly water, that further extend their flowering time. Even though agricultural areas also have supplementary resources, the main crop in flower in East Contra Costa County later in the season is maize, which is wind pollinated. There may be multiple impacts of exotic plants in urban areas. By filling the phenological flowering gap noted above, they may help attract even larger populations of bees into the urban landscape. In addition, bees in urban sites may be behaviorally more likely to visit non-native plants due to the increased encounters they have with novel plants. In agricultural and natural landscapes, a positive correlation between pollinator visitation and seed set is typical. Surprisingly in our system, in human-altered landscapes, higher total observed bee visitation did not result in higher proportions of seed set, as would be expected. In fact, urban areas, despite receiving the highest rates of native bee visitation, exhibited the lowest rates of seed set. Conversely, natural areas, which received the lowest amount of total bee visitation, had the highest rates of seed set. We suggest 2 possible explanations for this discrepancy between pollinator visitation and rates of seed set: 1) pollinator efficiency; and/or 2) the composition of the local flowering community. Depending on the plant, certain pollinator species are much more effective than others. For example, Osmia, Habropoda, and Apis, have been found to produce varying amounts of seed set as a result of a single visit to blueberry, but these results vary slightly depending on the blueberry variety. In the case of yellow starthistle, it is likely that the most frequent visitors are perhaps not the most efficient. When we directly compared average seed set at each site against visitation rates, we found a significant positive association with the medium hairy leg bees. The medium hairy leg bee morphotype includes those species which fall in both the Tribes Emphorini and Eucerini. Emphorini are known to largely be oligolectic , meaning they specialize on certain plant groups, which theory suggests would make them more efficient pollinators than generalists. The medium hairy leg bee morphotype was not significantly associated with any of the land use typesIt was also the only group that was observed most frequently during morning sampling, perhaps reflecting a difference in when yellow starthistle is most receptive to pollination. Despite the overwhelming abundance of honey bees in agriculture areas, we did not observe higher seed set in those regions, consistent with the observation that honey bees can be poorer pollinators than other species.It is also important to note that this study used a morphotype classification, and there may be multiple species that fit within the same morphotype that provide varying degrees of pollination services.