Remote quantification of the micro-scale vegetation structural features influencing a rare forest antelope's habitat selection
Lyndon
D.
Estes, RSCF/U. Virginia, lde2c@virginia.edu
(Presenting)
Adam
G.
Mwangi, Moi University/Kenya Wildlife Service, adamgm2003@yahoo.com
Gregory
S.
Okin, UCLA, okin@geog.ucla.edu
Paul
R.
Reillo, RSCF, paulreillo@rarespecies.org
Herman
H.
Shugart, U. Virginia, hhs@virginia.edu
1.The mountain bongo (Tragelaphus euryceros isaaci) is a highly endangered antelope endemic to Kenyan montane forests. Succesfully conserving the bongo depends on understanding its habitat selection and distribution using predictive models. Obtaining the data needed for such models is difficult given the bongo's rarity and the challenging terrain. Remote sensing (RS) is therefore required for characterizing important habitat features.
2.Bongo select habitat at the micro-scale of 0.04 ha, where selection is determined by understorey characteristics, and at the patch scale (~20 ha), where canopy characteristics determine the broader availability of browse and cover (Estes et al, 2008). Microhabitat features are only detectable from the ground, while the patch-scale features are best detected remotely.
3.Since both scales are important in shaping bongo habitat, and since distribution models depend on spatially continuous predictor variables, the need remains to quantify micro-scale features using RS data. In order to better characterize vegetation structure remotely, we re-expressed field variables using structural complexity indices (SCIs). SCIs were derived using: 1) principal components analysis, 2) factor analysis, and 3) a manual, additive technique (McElhinney et al, 2006).
4.To test the detectability of SCIs, we derived predictors from RS data using 1) a four end-member spectral mixture analysis (SMA) of ASTER data to obtain sub-pixel estimates of green vegetation, non-photosynthetic vegetation, soil, and shade abundance, and 2) mean and standard deviation texture analysis of the ASTER visual and near infrared bands. We derived additional predictors from an SRTM DEM resampled to 30 m. We will assess the utility of the SCIs through least squares and logistic regression models to determine 1) the accuracy with SCI values are predicted by RS variables, and 2) the ability of RS-predicted SCI values to discriminate between bongo and non-bongo habitat.
NASA Carbon Cycle & Ecosystems Active Awards Represented by this Poster: