Ecogeographical variables
A set of criteria was developed to predict raptor habitat suitability
which included choosing variables that were potentially useful for
raptors’ distribution. The Worldclim database version 2.1 (Fick and
Hijmans, 2017; https://www.worldclim.com/current) provided 19
bioclimatic variables. Because of their direct effects on species
distribution, climate variables are commonly employed in habitat
modeling . The Shuttle Radar Topographic Mission’s digital elevation
model (SRTM DEM, opendata.rcmrd.org/datasets/) was used to extract
digital elevation data for Kenya, slope and aspect were derived from the
DEM, and the topographic roughness index calculated as the surface area
ratio, which is also derived from the DEM . Data for the Normalized
Difference Vegetation Index (NDVI) (2013-2020) were obtained from NOAA,
as were observations from the Advanced Very High Resolution Radiometer
(AVHRR) and Visible Infrared Imaging Radiometer Suite (VIIRS) satellites
(https://www.star.nesdis.noaa.gov/smcd/emb/vci/VH/vh_browseByCountry.php).
The Human Influence Index (HII), which best represents anthropogenic
impacts spanning the years 1995 to 2004, was derived from Last of the
Wild v2
(https://sedac.ciesin.columbia.edu/data/collection/wildareas-v2/).
ArcGIS was used to rasterize all predictions at a spatial extent n of
about 1 km (Version 10.5). We used the ‘usdm ’ package in R to
carry out a variance inflation factor stepwise procedure to decrease
multicollinearity in predictor variables . Variables with variance
inflation factors greater than 10 were eliminated. As a result, we only
kept the 12 best-fitting covariates based on the raptors’ ecological
requirements from an initial set of 25 variables.