Appendix A (Table A1). Predictor variables and scales of association used to model the multi-scale response of avifauna to habitat, prescribed fire, and fire severity at four study sites in Arizona and New Mexico, USA.

Variable name

Scale of association

Level of measurement

Description

FAC_1

Point

Continuous

First of three retained factors resulting from the habitat factor analysis of tree density, basal area mean and SD tree height, ponderosa pine dominance, large tree richness, large snag count and decay, and canopy cover data collected on four 50-m transects around a point prior to treatment. Points with higher values for FAC_1 regression scores were best characterized by increased decay levels in the larger snags, higher stem densities, and greater tree species richness.

FAC_2

Point

Continuous

Second of three retained factors resulting from the habitat factor analysis of tree density, basal area mean and SD tree height, ponderosa pine dominance, large tree richness, large snag count and decay, and canopy cover data collected on four 50-m transects around a point prior to treatment. Points with higher values for FAC_2 regression scores were dominated by more open stands of older ponderosa pine.

FAC_3

Point

Continuous

Third of three retained factors resulting from the habitat factor analysis of tree density, basal area mean and SD tree height, ponderosa pine dominance, large tree richness, large snag count and decay, and canopy cover data collected on four 50-m transects around a point prior to treatment. Points with higher values for FAC_3 regression scores best characterized by high values for canopy closure and basal area.

PatchType

Patch

Ordinal

Three patch types derived from pre-treatment digital forest structure data (canopy cover, stem density, and basal area) synthesized using a fuzzy classification algorithm in BoundarySeer®. Types 1, 2, and 3 defined as “open,” “less dense,” and “more dense” respectively. Used nearest neighbor function in a GIS to summarize “majority” value within a 100-m radius of each cell on a unit.

PatchVariety

Patch

Ordinal

Count of the total number of different PatchType values occurring within a 100-m radius of each cell on a unit. Used nearest neighbor function in a GIS to summarize this “variety” value.

Elevation

Patch

Continuous

Used USGS digital elevation models and a GIS to derive mean value for elevation at each cell on a unit using a nearest neighbor function (focal mean) and a circular moving window with a 100-m radius. Standardized to mean zero and unit variance.

Slope

Patch

Continuous

Used USGS digital elevation models and a GIS to derive mean value for slope at each cell on a unit using a nearest neighbor function (focal mean) and a circular moving window with a 100-m radius. Standardized to mean zero and unit variance.

Fire severity level

Point

Ordinal

For models of avian response to fire severity level, evaluated three functional forms of this response using a simple linear model of the ordinal levels of fire severity and two transformations: log-normal and second-degree polynomial.

Treatment

Unit

Ordinal

Quantitative descriptor of “control” or “treatment” unit using 0 and 1, respectively.

Site

Site

Nominal

Qualitative descriptor of site (ASNF, CNF, GNF, KNF) used as random effect in all mixed models.