Quantifying patterns and spatial contagion of urban pollinator gardening using Google Street View

Background/Question/Methods

Urban landscapes are novel and heavily-human modified ecosystems that can nevertheless harbor surprisingly high biodiversity, leading to increasing interest in conservation in these areas. Pollinator gardening is one promising avenue for urban conservation because it provisions valuable nectar and pollen resources for ecologically-crucial pollinator populations, which are feared to be in decline. However, we don't know much about this habitat type's current distribution, even basic things like how much there is currently. My research assesses the prevalence of pollinator gardening on the urban landscape and tests for evidence of social contagion, the phenomenon in which nearby individuals tend to mimic each other’s choices. Under social contagion we predict pollinator gardens will be clustered together spatially and be most clustered at intermediate levels of overall pollinator garden density. We selected 10 cities in the Boston area of varying income and lot sizes, then manually scored Google Street View imagery of all residential addresses for the presence of pollinator-friendly, diverse flower gardens. We then visited a stratified random subset of 500 addresses to assess flowering species diversity to ground truth these imagery scores. We quantified differences in probability of gardening and the prevalence of flower gardening using binomial GLMs.

Results/Conclusions

We hand-classified 81,074 Street View images into categories based on dominant front yard cover (pollinator gardens, weedy yards, other vegetation, grass lawn, or impermeable). Overall prevalence of pollinator gardening was low: over the 10 cities, 1.6% of addresses were scored as pollinator gardens, but the proportion of gardens varied significantly across cities (range 1.0% - 2.8%, likelihood-ratio test on binomial GLM χ2 = 99.9, df = 9, p = 0). Street View scores were effective at capturing variation in pollinator-visited plant diversity in urban yards: the yards scored as pollinator gardens on Street View had an average of >2 times more pollinator-friendly species blooming (8.7 vs. 3.8 species in flower), in spite of the coarse image scoring scale necessitated by image resolution, which made explicitly identifying species impossible. Preliminary analyses suggest that pollinator gardening is strongly spatially clustered within cities and may also be associated with income and negatively associated with lot size. These patterns are consistent with the hypothesis that the adoption of pollinator gardening is partially driven by social contagion, but that it is currently not a very widespread practice in this region. Our mapping method allows us to quantify prevalence and spread of this socially-driven conservation practice.