Supplementary Information Schuetz et al.
Detailed Methods
We entered common names of bird species individually into Google Trends, downloaded corresponding csv files, and collated normalized search indices for all species for each of the 50 states from 2004-2013. We restricted our focus to resident birds because occurrences and detection probabilities of migrant species vary considerably across space and time, significantly complicating analyses of potential relationships between bird densities and Internet search activity. Not all resident species were queried frequently enough for Google Trends to summarize search activity and those species were removed from further consideration.
To generate density estimates for each bird in each state, we divided state population estimates by the land area of that state. We then normalized values for each species relative to the state in which it reached its maximum density. So, for each species, the state with the highest density of birds received a value of 100. Values for all other states were expressed as a percentage of that maximum.
We decided to use mixed effects models to analyze the data for two reasons: 1) treating species as a fixed effect would require a large number of parameters to be estimated; and 2) our primary goal was to assess the degree to which search activity was shaped by bird distributions (and other factors), not to estimate the effects of individual species or states on the relationship between bird density and Internet search activity.
When implementing linear mixed-effects models, we included random intercept and slope effects with species as the grouping variable and bird density as the random slope variable. To account for repeated observations from each state, we also included a random intercept effect with US state as the grouping variable. We generated a model set containing all additive combinations of fixed effects (16 models) using the lme4 and MuMIn packages in R (Pinheiro et al. 2013; Barton 2013; R Core Team 2013). Parameter estimates were model-averaged to develop a consensus model describing the relationships between predictor and response variables.
We repeated the analysis on a subset of the data in order to remove the confounding effects of searches for non-bird topics from our investigation. To establish which species names might be confounded with non-bird interests and search activity, we performed a Google search for each of the 68 resident bird species. If any of the top 20 search results included a site that was not directly related to the species or birding, we decided that searches for the species name might be confounded with searches on unrelated topics. The number of confounding results varied considerably among species but, for simplicity, we removed all species from the secondary analysis whose names might be searched in contexts unrelated to birds (see Table S2). We also repeated our analysis using only game birds to focus attention on species that might be queried by a particular segment of the population (i.e., hunters) (Table S3).
Diagnostic and qq plots (Figures S1-S2) indicate that the assumptions of linear mixed-effects modeling were met for the main analysis conducted in this study (all resident birds). More specifically, variance in the residuals did not change markedly with the value of the response variable, the residuals themselves were close to normally distributed, and there was no clear relationship between the residuals and each of the predictor variables. The pattern of decreasing residuals with increasing fitted values is due to the nature of the response variable, an index that ranged from 0-100. We chose not to transform the response because the other diagnostics were satisfactory and we did not want to complicate interpretation of the results. The kink in the qq plots is also a product of the response variable. Diagnostics for the other analyses done in this study (widely-distributed species, by region, birds without confounded names, game birds) were similar and suggested that the analytical methods selected were appropriate.
References
Barton K (2013) MuMIn: Multi-model inference. R package version 1.9.13.
http://CRAN.R-project.org/package=MuMIn.
Pinheiro J, Bates D, DebRoy S, Sarkar D, R Development Core Team (2013) nlme: Linear and Nonlinear
Mixed Effects Models. R package version 3.1-109.
R Core Team (2013) R: A language and environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria. URL http://www.R-project.org/.
Table S1. List of taxa included in the full analysis.
Acorn Woodpecker / Melanerpes formicivorus
American Dipper / Cinclus mexicanus
Anna's Hummingbird / Calypte anna
Arizona Woodpecker / Picoides arizonae
Barn Owl / Tyto alba
Barred Owl / Strix varia
Black-capped Chickadee / Poecile atricapilla
Black Phoebe / Sayornis nigricans
Black Vulture / Coragyps atrata
Blue Jay / Cyanocitta cristata
Bushtit / Psaltriparus minimus
Cactus Wren / Campylorhynchus brunneicapillus
California Quail / Callipepla californica
California Thrasher / Toxostoma redivivum
California Towhee / Melozone crissalis
Canyon Wren / Catherpes mexicanus
Carolina Chickadee / Poecile carolinensis
Carolina Wren / Thryothorus ludovicianus
Chukar / Alectoris chukar
Common Raven / Corvus corax
Crested Caracara / Caracara cheriway
Downy Woodpecker / Picoides pubescens
Eastern Screech Owl / Megascops asio
Eurasian Collared Dove / Streptopelia decaocto
European Starling / Sturnus vulgaris
Gambel's Quail / Callipepla gambelii
Gila Woodpecker / Melanerpes uropygialis
Gray Jay / Perisoreus canadensis
Great Gray Owl / Strix nebulosa
Great Horned Owl / Bubo virginianus
Great Kiskadee / Pitangus sulphuratus
Great-tailed Grackle / Quiscalus mexicanus
Greater Roadrunner / Geococcyx californianus
Hairy Woodpecker / Picoides villosus
House Finch / Carpodacus mexicanus
House Sparrow / Passer domesticus
Inca Dove / Scardafella inca
Mexican Jay / Aphelocoma wollweberi
Mountain Chickadee / Poecile gambeli
Mountain Quail / Oreortyx pictus
Northern Cardinal / Cardinalis cardinalis
Northern Mockingbird / Mimulus guttatus
Oak Titmouse / Baeolophus inornatus
Phainopepla / Phainopepla nitens
Pileated Woodpecker / Dryocopus pileatus
Pinyon Jay / Gymnorhinus cyanocephalus
Pyrrhuloxia / Cardinalis sinuatus
Red-bellied Woodpecker / Melanerpes carolinus
Red-cockaded Woodpecker / Picoides borealis
Red-shouldered Hawk / Buteo lineatus
Rock Pigeon / Columba livia
Ruffed Grouse / Bonasa umbellus
Scaled Quail / Callipepla squamata
Seaside Sparrow / Ammodramus maritimus
Snail Kite / Rostrhamus sociabilis
Spruce Grouse / Falcipennis canadensis
Steller's Jay / Cyanocitta stelleri
Tufted Titmouse / Baeolophus bicolor
Verdin / Auriparus flaviceps
Western Screech Owl / Megascops kennicottii
Western Scrub Jay / Aphelocoma californica
White-breasted Nuthatch / Sitta carolinensis
White-tailed Hawk / Buteo albicaudatus
White-tipped Dove / Leptotila verreauxi
Wild Turkey / Meleagris gallopavo
Willow Ptarmigan / Lagopus lagopus
Wrentit / Chamaea fasciata
Yellow-billed Magpie / Pica nuttalli
Table S2. Species whose names appear in search results unrelated to birds.
Species / Number of non-bird results out of 20Barred Owl / 1
Common Raven / 1
Ruffed Grouse / 1
Carolina Wren / 2
Great Horned Owl / 2
Cactus Wren / 3
Chukar / 3
Wild Turkey / 7
Canyon Wren / 8
Barn Owl / 9
Blue Jay / 9
Verdin / 15
Table S3. List of species classified as game birds.
Game speciesCalifornia Quail
Chukar
Eurasian Collared Dove
Gambel's Quail
Mountain Quail
Ruffed Grouse
Scaled Quail
Spruce Grouse
White-tipped Dove
Wild Turkey
Willow Ptarmigan
Figure S1. Diagnostic plots for the analysis that included all resident bird species.
Figure S2. QQ plot for the analysis that included all resident bird species.
Table S4. Model-averaged parameter estimates for the analysis that excluded confounding species.
Estimate / SE / Lower 95% CI / Upper 95% CIIntercept / 22.66 / 10.51 / 2.05 / 43.26
Bird density / 0.52 / 0.05 / 0.42 / 0.61
Hunters / -0.35 / 0.16 / -0.67 / -0.03
Inactivity / -0.56 / 0.29 / -1.13 / 0.01
Birders / -0.25 / 0.19 / -0.63 / 0.12
Table S5. Model-averaged parameter estimates for the analysis of game species.
Estimate / SE / Lower 95% CI / Upper 95% CIIntercept / 6.85 / 4.07 / -1.13 / 14.82
Bird density / 0.68 / 0.10 / 0.48 / 0.88
Hunters / -0.02 / 0.09 / -0.20 / 0.16
Inactivity / -0.06 / 0.16 / -0.37 / 0.24
Birders / -0.02 / 0.09 / -0.20 / 0.15
Table S6. List of taxa found in more than half of all states.
Common Name / Scientific NameBarred Owl / Strix varia
Black-capped Chickadee / Poecile atricapilla
Blue Jay / Cyanocitta cristata
Carolina Wren / Thryothorus ludovicianus
Common Raven / Corvus corax
Downy Woodpecker / Picoides pubescens
Eastern Screech Owl / Megascops asio
European Starling / Sturnus vulgaris
Gray Jay / Perisoreus canadensis
Great Kiskadee / Pitangus sulphuratus
House Finch / Carpodacus mexicanus
House Sparrow / Passer domesticus
Inca Dove / Scardafella inca
Northern Mockingbird / Mimulus guttatus
Oak Titmouse / Baeolophus inornatus
Pinyon Jay / Gymnorhinus cyanocephalus
Red-cockaded Woodpecker / Picoides borealis
Rock Pigeon / Columba livia
Ruffed Grouse / Bonasa umbellus
Scaled Quail / Callipepla squamata
Verdin / Auriparus flaviceps
White-tailed Hawk / Buteo albicaudatus
Willow Ptarmigan / Lagopus lagopus
Figure S3.The four flyways as defined by the National Audubon Society for regional analyses. Bird density data were unavailable for Hawaii.
Supplementary Information Schuetz et al.
Figure S4.Northern and southern states as classified for regional analyses. Bird density data were unavailable for Hawaii.
Supplementary Information Schuetz et al.
Table S7. Model-averaged parameter estimates for the analysis performed on widely-distributed species.
Estimate / SE / Lower 95% CI / Upper 95% CIIntercept / 25.26 / 11.04 / 3.63 / 46.90
Bird density / 0.42 / 0.07 / 0.28 / 0.56
Hunters / -0.45 / 0.23 / -0.91 / 0.01
Inactivity / -0.29 / 0.37 / -1.01 / 0.42
Birders / 0.01 / 0.22 / -0.43 / 0.45
Table S8. Model-averaged parameter estimates for the regional analyses.
Estimate / SE / Lower 95% CI / Upper 95% CI
Intercept / 63.44 / 37.02 / -9.12 / 136.00
Bird density / 0.47 / 0.06 / 0.34 / 0.59
Hunters / -0.65 / 0.50 / -1.62 / 0.32
Inactivity / -0.61 / 1.33 / -3.21 / 2.00
Birders / -0.89 / 0.74 / -2.34 / 0.57
Central Flyway
Estimate / SE / Lower 95% CI / Upper 95% CI
Intercept / 14.59 / 24.77 / -33.95 / 63.13
Bird density / 0.26 / 0.04 / 0.17 / 0.35
Hunters / -0.30 / 0.50 / -1.28 / 0.68
Inactivity / 0.29 / 0.89 / -1.46 / 2.04
Birders / -0.35 / 0.47 / -1.27 / 0.56
Mississippi Flyway
Estimate / SE / Lower 95% CI / Upper 95% CI
Intercept / 69.17 / 12.90 / 43.90 / 94.45
Bird density / 0.22 / 0.07 / 0.08 / 0.35
Hunters / -0.81 / 0.26 / -1.32 / -0.29
Inactivity / -1.54 / 0.31 / -2.15 / -0.93
Birders / 0.08 / 0.24 / -0.40 / 0.55
Atlantic Flyway
Estimate / SE / Lower 95% CI / Upper 95% CI
Intercept / -2.05 / 43.67 / -87.64 / 83.54
Bird density / 0.38 / 0.06 / 0.27 / 0.49
Hunters / -0.68 / 0.53 / -1.72 / 0.35
Inactivity / 0.90 / 1.54 / -2.11 / 3.92
Birders / 0.10 / 0.57 / -1.03 / 1.22
Northern States
Estimate / SE / Lower 95% CI / Upper 95% CI
Intercept / 29.60 / 17.93 / -5.54 / 64.74
Bird density / 0.30 / 0.05 / 0.21 / 0.40
Hunters / -0.32 / 0.20 / -0.71 / 0.07
Inactivity / -0.56 / 0.59 / -1.73 / 0.60
Birders / -0.10 / 0.25 / -0.59 / 0.39
Southern States
Estimate / SE / Lower 95% CI / Upper 95% CI
Intercept / 30.19 / 12.59 / 3.63 / 46.90
Bird density / 0.49 / 0.04 / 0.28 / 0.56
Hunters / -0.74 / 0.45 / -0.91 / 0.01
Inactivity / -0.20 / 0.50 / -1.01 / 0.42
Birders / -0.16 / 0.37 / -0.43 / 0.45