Influences of horizontal and vertical aspects of land cover and their interactions with regional factors on patterns of avian species-richness

Figures & data

Table 1. Classification of land-cover classes based on landscape vegetative strata

Landscape vegetation class	NLCD classes
Strata 0	Open water
	Perennial ice/snow
	High-intensity residential
	Commercial/industrial/transportation
	Bare rock/sand/clay
	Quarries/strip mines/gravel pits
	Transitional
Strata 1	Low-intensity residential
	Grasslands/herbaceous
	Pasture/hay
	Row crops
	Small grains
	Fallow
	Urban/recreational grasses
	Emergent herbaceous wetlands
Strata 2	Shrubland
	Orchards/vineyards/other
Strata 3	Deciduous forest
	Evergreen forest
	Mixed forest
	Woody wetlands

Notes: “Strata 0” refers to land-cover classes with indeterminate strata levels, and the other three classes refer to the number of vegetative strata associated with their respective land-cover classes.

Figure 1. Map of 1992 North American breeding bird survey routes color-coded by species-richness estimates, with values divided into five groups at natural breaks.

Notes: Longitude and latitude coordinates identified in the map correspond to splitting points in the regression tree (see Figure 3; longitude = −97.44, latitude = 35.98). Land and water base layers were generated using free spatial data from Natural Earth. Source: http://www.naturalearth.com/.

Figure 2. Scatter plots of species-richness estimates against latitude, longitude, mean elevation, standard deviation of elevation, land-cover diversity, and proportions of the four strata classes present within 5 km buffers around each BBS route.

Table 2. Confidence intervals (95% on α=0.05) for coefficient estimates of listed variables, significance levels of the same, and the last three digits (all numbers began with 16) of model-wise AIC values for each of the multiple-regression models examined (10 models total)

Regression set	Variable	CI βi	p-value	AIC
Null model	Latitude	0.69 ± 0.12	<2e-15	902
	Longitude	0.42 ± 0.05	<2e-15
	Elevation (mn)	−0.01 ± 0.00	<2e-15
	Elevation (sd)	0.04 ± 0.01	<2e-15
Geographic variables with single landscape feature	Diversity	11.24 ± 1.51	<2e-15	702
	Strata 0*	−9.46 ± 6.97	0.079	897
	Strata 1	−3.97 ± 2.24	0.005	892
	Strata 2	−0.15 ± 3.20	<2e-15	823
	Strata 3	12.91 ± 2.26	<2e-15	783
Combined landscape features	Diversity	12.22 ± 1.53	<2e-15	668
	Strata 0	−20.07 ± 6.742.25	e-08
	Diversity	11.21 ± 1.53	<2e-16	697
	Strata 1*	−2.91 ± 2.15	0.078
	Diversity	10.07 ± 1.55	<2e-15	667
	Strata 2	−9.82 ± 3.181.7	4e-08
	Diversity	10.01 ± 1.51	<2e-15	621
	Strata	310.34 ± 2.21	<2e-15

Notes: “Null model” refers to the model containing only geographic variables, “Geographic variables with single landscape feature” refers to the five models each containing the geographic variables and only 1 of the landscape features examined, and “Combined landscape features” refers to the four models containing land-cover diversity and proportion of one of the four strata classes together with the geographic variables. The values of the coefficients for latitude, longitude, mean elevation, and standard deviation of elevation varied with each model, with the coefficient for mean elevation not significantly different to 0 in the model with land-cover diversity and strata 3.

* Indicates that the estimated coefficient failed to meet the significance threshold required to reject the null hypothesis that its slope is not significantly different to 0.

Table 3. Tabular summary of regression-tree output depicting each node with the value of the splitting point, the number of data points on its right-hand (n), and their deviance and mean

Node	Splitting point	n	Deviance	Mean
1		2,087	629,800	62.18
. 2	lon < −97.4417	795	234,200	49.68
4	lcd < 0.96	293	69,180	40.88
8*	ps3 < 0.008	133	20,350	34.37
9*	ps3 > 0.008	160	38,520	46.29
5	lcd > 0.96	502	129,100	54.81
10*	lcd < 1.54	371	88,690	52.21
11*	lcd > 1.54	131	30,750	62.19
3	lon > −97.44	1,292	194,900	69.87
6*	ps3 < 0.25	295	39,620	63.41
7	ps3 > 0.25	997	139,300	71.78
14*	lat < 35.975	307	29,500	65.77
15*	lat > 35.975	690	93,760	74.46

Notes: The abbreviation “lat” refers to latitude, “lon” to longitude, lcd to land-cover (horizontal) diversity, and psx to proportion of (vertical) strata class x in the buffer. Node 1 refers to the entire dataset.

* Indicates a terminal node.

Figure 3. Plot of the regression tree generated using all geographic and land-cover variables together.

Notes: The left side of the split indicates that the conditional is true, and the right side that it is false. The vertical length of branches indicates the amount of deviance reduction. Variable abbreviations as follows: lon = longitude, lcd = land-cover diversity, ps3 = proportion of cells in strata-class 3, and lat = latitude.