A classification approach to categorize urbanized landscapes in California and trait variation in Maltese starthistle

There is increasing evidence that the traits of organisms can differ in urban environments, but defining what makes an environment "urban" is difficult. There are many variables associated with increasing human impacts that might be associated with urbanization, and this makes it challenging to identify both how traits vary across diverse urbanized landscapes and the variables that might drive that variation. To better define the mosaic of urban environmental heterogeneity and identify similar environments that are comparable across its complexity, we developed a multivariate landscape classification framework and applied it to classification of land area in the state of California, USA. We then asked if these multivariate Environmental Zones were better able to explain trait variation than univariate urban versus nonurban dichotomies or gradients, using a case study of the invasive, annual plant Maltese starthistle (Centaurea melitensis), commonly found throughout much of California. We used a hierarchical cluster analysis to group 7829 government census tracts into Environmental Zones based on a set of 19 independent environmental characteristics, including climate, land cover, pollution, and socioeconomic variables. This analysis identified nine major Environmental Zones, which were differentiated based upon complex combinations of variables that did not align with urban versus natural dichotomies or gradients. Environmental Zones also occurred as mosaics of many zones within cities and differed in their relative abundance between cities, reflecting complex urban landscapes unique to each area. Seeds from 17 populations of Maltese starthistle were collected from six of the Environmental Zones, including two more heavily urbanized zones, and four more natural/agricultural. Seeds were grown in greenhouse conditions, and eight vegetative traits were measured. No trait differed significantly between qualitatively urban versus nonurban sites, or across a quantitative gradient of impervious surface area, but four traits differed according to Environmental Zones (length of longest leaf, specific leaf area [SLA], root diameter, and the number of flowerheads). Traits differed between the more urbanized zones, as well as among the more natural zones. Our results reveal that more complex multivariate classifications of the urban mosaic can identify similar, comparable environments across complex landscapes and better explain trait variation in organisms navigating urbanized environments.