The importance of street trees to urban avifauna

Abstract Street trees are public resources planted in a municipality’s right‐of‐way and are a considerable component of urban forests throughout the world. Street trees provide numerous benefits to people. However, many metropolitan areas have a poor understanding of the value of street trees to wildlife, which presents a gap in our knowledge of conservation in urban ecosystems. Greater Los Angeles (LA) is a global city harboring one of the most diverse and extensive urban forests on the planet. The vast majority of the urban forest is nonnative in geographic origin, planted throughout LA following the influx of irrigated water in the early 1900s. In addition to its extensive urban forest, LA is home to a high diversity of birds, which utilize the metropolis throughout the annual cycle. The cover of the urban forest, and likely street trees, varies dramatically across a socioeconomic gradient. However, it is unknown how this variability influences avian communities. To understand the importance of street trees to urban avifauna, we documented foraging behavior by birds on native and nonnative street trees across a socioeconomic gradient throughout LA. Affluent communities harbored a unique composition of street trees, including denser and larger trees than lower‐income communities, which in turn, attracted nearly five times the density of feeding birds. Foraging birds strongly preferred two native street‐tree species as feeding substrates, the coast live oak (Quercus agrifolia) and the California sycamore (Platanus racemosa), and a handful of nonnative tree species, including the Chinese elm (Ulmus parvifolia), the carrotwood (Cupaniopsis anacardioides), and the southern live oak (Quercus virginiana), in greater proportion than their availability throughout the cityscape (two to three times their availability). Eighty‐three percent of street‐tree species (n = 108, total) were used in a lower proportion than their availability by feeding birds, and nearly all were nonnative in origin. Our findings highlight the positive influence of street trees on urban avifauna. In particular, our results suggest that improved street‐tree management in lower‐income communities would likely positively benefit birds. Further, our study provides support for the high value of native street‐tree species and select nonnative species as important habitat for feeding birds.

. Summaries of migratory study bird foraging observations. Table S2. Summaries of year-round study bird foraging observations. Table S3. Summaries of street tree density and size. Table S4. Street-tree importance values and feeding bird observations for all street tree species. Table S5. Summaries of migratory bird foraging behavior on nine common tree species.  2 Appendix S1

Spatial autocorrelation
A critical concern when designing a study across a spatially segregated gradient, such as the socioeconomic gradient in LA, are patterns of autocorrelation among survey locations (Legendre 1993). Issues with autocorrelation can violate the assumption of independence, which is a key assumption of many statistical analyses (Zuur et al. 2011). Before analysis, we tested for patterns of spatial dependence among our survey locations by computing a semivariogram analysis (Legendre and Fortin 1989) using the 'geoR' package in R (Ribeiro Jr and Diggle 2018).
The semivariogram analysis calculated the degree of dissimilarity of the residuals of bird and tree variables (see statistical analysis for details on variables) between survey locations as a function of distance (Legendre and Fortin 1989). There were no issues with spatial dependence of our survey locations based on the semivariogram analysis, and therefore, we considered all survey locations as independent (Appendix S1: Fig. S1).

Flock-density estimation
Due to the flocking nature of a handful of year-round focal species (e.g., the Bushtit, Psaltriparus minimus, the Lesser Goldfinch, Spinus psaltria, and the House Finch, Haemorhous mexicanus), we opted to consider larger, homogenous flocks of these species as an n = 1. We employed the reduced method to avoid overinflating the ecological importance of a given tree on the movement and feeding patterns of a group of birds. To ascertain whether the 'reduced' method yielded similar results compared with including all individuals of the same species within a flock ('all'), we computed two analyses. First, we calculated a Spearman's rho (r) 3 correlation between the density of 'reduced' and 'all' year-round birds. The Spearman's r value was 0.81, p-value < 0.01 (Appendix S1: Fig. S2). Further, we computed an ANOVA analysis of both 'reduced' and 'all' density measures among the three income groups of our study. Our intention with the ANOVA analysis was to compare similarities in effect sizes between response variables. We found that the 'reduced' measure yielded an F-value of 5.182, and indicated strong significant differences (p-value = 0.011) of year-round bird density among income groups (Table   1, in manuscript). We found nearly identical effect-sizes for the 'all' measure, F-value of 5.182, p-value = 0.011. Our results support that our 'reduced' treatment of the year-round flock density values yields similar effects to the 'all' measure.

Street-tree phenology and foraging-bird behavior
To understand patterns in bird feeding depending on the phenology of tree species throughout the winter months, we documented the dominant phenophase of a street tree in which we completed a foraging observation. We created the following categories to describe tree phenology: (a) 'leafed out', full leaf canopy with < 25% of visible flowers, seeds, fruits, or senescing leaves; (b) 'seeding', > 75% of a tree's canopy containing visible seeds or nuts; (c) 'flowering', > 75% of a tree's canopy containing visible flowers; (d) 'fruiting', > 75% of canopy containing visible fleshy fruit; (e) 'leaf senescence', > 75% of canopy with senescing leaves, or (f) 'leafless', a winter deciduous or dead tree, with no visible live leaves.
In general, birds exhibited variable feeding patterns based on differences in phenology of street trees. When examining the dominant phenological stage of both native and non-native trees in which we observed migratory bird foraging, trees were primarily leafing (n = 192, 35.75%), followed by seeding (n = 138, 25.70%), flowering (n = 80, 14.90%), fruiting (n = 51, 4 9.50%), leaf senescence (n = 72, 13.41%), or leafless (n = 4, 0.74%). Of the 587 observed foraging observations and 2246 attack maneuvers of migratory birds, 47.06% of attacks were on leaf surfaces (n = 1057), 20.30% were aerial attacks (n = 456), 14.95% were on woody surfaces (n = 336), 12.37% were on flowers (n = 278), and 5.29% were on fruit surfaces (n = 119). We recorded migratory bird feeding in 49 other individual trees where we did not record treephenology stage. For year-round birds, we did not record detailed attack maneuvers on a tree substrate in which a bird was feeding, but we did record the dominant phenology stage of each tree in which we made a bird observation. Of the 351 observations, year-round birds primarily fed in trees that were seeding (n = 175, 54.86%), leafing (n = 64, 20.06%), flowering (n = 53, 20.06%), fruiting (n = 14, 4.39%), or leafless (n = 14, 4.39%). We recorded resident bird feeding in 32 other tree species where we did not record dominant tree phenology.
The foraging behavior of birds can be used as an indirect proxy for the availability of invertebrate prey items on trees species (Lovette and Holmes 1995). We found that the attack index of birds, which is defined as a ratio of the number of attacks to search maneuvers, standardized to a similar time of observation (Lovette and Holmes 1995, Oyugi et al. 2012, Wood et al. 2012, was similar among nine common tree species in which we had sufficient foraging data for analysis (Appendix S1: Table S5). We performed an ANOVA analysis to test whether there were differences in the mean attack rate among tree species, which revealed possible differences (F 8, 267 = 1.96, p-value = 0.05). However, a subsequent Tukey's HSD test yielded no significant differences in the attack index among tree species. Attacks, ignoring search efforts, were more variable among tree species (F 8, 267 = 3.31, p-value < 0.01, Appendix S1: Table S5). However, after computing a Tukey's HSD test, the only difference, based on a Bonferroni adjusted alpha value of 0.05/35 comparisons = 0.0014, was between the Camphor 5 Tree and the Italian Stone Pine (p-value = 0.0003). There were no differences in total search efforts among the nine common tree species (F 8, 267 = 1.71, p-value = 0.08).
Literature Cited Legendre, P. 1993 Table S1. Foraging behavioral observations of five migratory bird species observed feeding on public street trees during two winter field seasons along a socioeconomic gradient of low (< $53,219 medium household income), medium ($53,220 -$70,719), and high-income residential communities (> $70,720) throughout Greater Los Angeles. Numbers indicate the total number of foraging observations, whereas values in superscript represent the cumulative number of seconds we observed a bird foraging. Numbers in superscript within parentheses indicate total attacks and total searches. 7 Table S2. Foraging behavioral observations of five year-round bird species observed feeding on public street trees during two winter field seasons along a socioeconomic gradient of low (< $53,219 medium household income), medium ($53,220 -$70,719), and high-income residential communities (> $70,720) throughout Greater Los Angeles. Numbers indicate the total number of foraging observations. We did not monitor individual foraging behavior of year-round species as observations were often of a stationary bird feeding. Angeles study locations. We define feeding as an observed feeding maneuver on a tree substrate (i.e., an attack directed at the bark, leaf, bud, flower, fruit, seed, or aerial maneuver within the limits of the tree canopy). Feeding proportion values higher than street tree IVs suggest foraging preference, whereas feeding proportion values lower than street tree IVs suggest foraging avoidance by migratory or year-round birds.   Table S5. Summary of migratory bird foraging behavior on nine common street-tree species in which we had sufficient foraging data (> 10 unique foraging observations that were > 30 sec in continuous observation). We display the number (n), total and mean observation time in seconds, the mean number of attacks and searches by migratory birds on each tree standardized per minute, and the attack index. The attack index -also referred to as the attack rate -is a ratio of attacks to searches, and is a measure of foraging  Figure S1. Semivariograms of the residuals of the observed number of feeding migratory birds (km), the observed number of feeding year-round birds (km), street-tree richness (km), Shannon Diversity of street trees, street-tree density (km), and street-tree basal area m 2 (km) plotted as a function of the spatial arrangement of 36 survey locations. The semivariogram analysis bins residuals from residential survey locations that were similar and the semivariance at each lag is  densities of year-round birds. 'Reduced' was a method where we treated all individuals of the same species within a large flock as an n = 1. We employed the reduced method to avoid overinflating the ecological importance of a given tree on the movement and feeding patterns of a group of birds. 'All' considered every individual bird within homogenous flocks. We standardized the tallies to 1-km of walking route, and thus refer to the numbers as density measures. The Spearman's r value of 0.81, p-value < 0.01 indicated a high correlation between both the reduced and all measures, suggesting both metrics yield similar effects in analyses.