DRECP species distribution model output for White-margined beardtongue
(Penstemon albomarginatus ).
Patrick McIntyre and Kara Moore of UC Davis generated predictions of
rare plant habitat occupancy from Maxent models using an approach designed to
maximize model ability to identify new occurrences in the field.
A three-stage approach was used to build, assess, and finalize
distribution models for focal taxa. First, the team built preliminary models at
270 m for each species. Second, field surveys were conducted based on
preliminary model predictions. Third, final models were built based on all
occurrences, including field data, and assessed for potential biases and model
fit. Models were evaluated for their ability to distinguish occurrences from
non-occurrences within a 20 km buffer around the known species range using cross
validated unadjusted AUC scores and calibrated AUC scores based on a pairwise
geographic distance correction designed to overcome spatial sorting bias.
Both continuous probability surfaces and binary layers are available
for each species modeled. Binary layers depicting predicted suitable habitat
were derived using a threshold maximizing the sum of sensitivity and specificity
(max SSS).
For more detailed information on environmental variables, methods, and
model diagnostics, please refer to the DRECP Working Group supporting document
"UCD_McIntrye and Moore Maxent Models for DRECP".
For Penstemon albomarginatus, Max Sensitivity + Specifity threshold =
0.43; AUC-cross validation = 0.865; pwdAUC = 0.741.