Transcript Future KBA Identification

Candidate KBA Identification:
Modeling Techniques for Field
Survey Prioritization
• Species Distribution Modeling: approximation of
species ecological niche projected into geographic
space
– realized niche may be smaller than fundamental or
“theoretical” niche
– due to many possible factors (such as geographic
barriers to dispersal, biotic interactions, and human
modification of the environment), few species occupy
all areas that satisfy their niche requirements
Limitations of Species Modeling
for KBA Delineation
• Presence data rarely accompanied by absence data
• Models often overestimate species extent (errors
of commission), which may lead to “protection”
where a target species does not actually occur
• Environmental data associated with samples may
not fully represent a species’ fundamental niche
Problems with presence-only
species data
•
•
•
•
Sampling bias
False negatives
Spatial auto-correlation
Variation in sampling intensity and sampling
methods used
• Errors in occurrence locality
• Errors in the recording associated environmental
variables
Environmental Variables
• Should be both temporal and spatial (scale) relationships
between variables and species requirements.
– climatic variables such as temperature and precipitation are
appropriate at global and meso-scales
– topographic variables (e.g., elevation and aspect) likely affect
species distributions at meso- and topo-scales
– land-cover variables (e.g., percent canopy cover) influence species
distributions at the micro-scale
– certain variables generalize well over large, regional scales
(bioclimatic and soil-type); others do not (elevation and latitude)
Models
• BIOCLIM
• DOMAIN
• Generalized Linear Models (GLM) and Generalized
Additive Models (GAM)
• Genetic Algorithm for Rule-set Prediction (GARP / OMGARP)
• Maximum Entropy (Maxent)
Model Comparison
AUC
Eliith et al. 2006. Novel methods improve prediction of species’ distributions from occurrence data. ECOGRAPHY vol 29.
Model Comparison
Darker areas represent greater inter-model agreement; circles
represent areas of over-estimation of ecological niche distribution.
Raxworthy et al. 2003. Predicting distributions of known and unknown reptile species in Madagascar. Nature vol 426
Additional Prioritization Options
for Field Surveys
• Identification of unstudied areas of greatest potential
biological diversity (via beta diversity modeling)
• Areas that have experienced highest rate and extent of
habitat loss
• Areas vulnerable to future threats (e.g., scenarios models
of infrastructure development)
– Areas of greatest climate change risk (El Niño and long-term)