Anthracothorax nigricollis

NOTES

To elaborate our models we removed uncertain localities such as 'Bogota skins' and all records with coordinates laying down in sites with estimated elevations above 1,900 m. Equally, we removed all records placed off the continent previous to modelling.

A first habitat suitability model generated in Maxent using all data showed very poor predictions in the Orinoco and Amazon regions. Therefore, we decided to model separately regions west and east of the Andes. The new habitat suitability model generated in Maxent west of the Andes showed very good predictions although probably with some omission error in the Caribbean. Also this same model showed some areas suitable for this hummingbird in the Pacific slope, for some of these areas there are no known records yet. However, they were left in our potential distribution map since they seem to agree with deforested areas where possibly the species has arrived recently and can point further areas where it might continue expanding its distribution in the Pacific. The habitat suitability model generated in Maxent east of the Andes showed a few areas suitable in climatic terms for this species west of the Andes and better predictions east of the Andes than when used all the data, although still with some degree of omission. Areas predicted west of the Andes by the east model and vice versa were removed previously to add up both models to have complete predictions from each one in the specified regions.

According to HBW Alive (2016) apparently records of subspecies iridescens from west Ecuador and northwest Peru in the high Cauca Valley are erroneous. There are 15 specimens determined as iridescens in BioMap from Valle del Cauca and Antioquia. This presents a taxonomic ambiguity since these specimens possibly represent a different variant to the nominate subspecies or variation within it. Nonetheless, this is pending further research.

MODEL METADATA

West model

Regularized training gain is 1.352, training AUC is 0.933, unregularized training gain is 1.709.

Algorithm terminated after 2000 iterations (79 seconds).

The follow settings were used during the run:

206 presence records used for training.

10205 points used to determine the Maxent distribution (background points and presence points).

Environmental layers used (all continuous): bio10co bio11co bio12co bio13co bio14co bio15co bio16co bio17co bio18co bio19co bio1co bio2co bio3co bio4co bio5co bio6co bio7co bio8co bio9co

Regularization values: linear/quadratic/product: 0.050, categorical: 0.250, threshold: 1.000, hinge: 0.500

Feature types used: hinge product linear threshold quadratic

responsecurves: true

jackknife: true

maximumiterations: 2000

East Model

Regularized training gain is 0.760, training AUC is 0.912, unregularized training gain is 1.411.

Algorithm converged after 960 iterations (25 seconds).

The follow settings were used during the run:

34 presence records used for training.

10033 points used to determine the Maxent distribution (background points and presence points).

Environmental layers used (all continuous): bio10co bio11co bio12co bio13co bio14co bio15co bio16co bio17co bio18co bio19co bio1co bio2co bio3co bio4co bio5co bio6co bio7co bio8co bio9co

Regularization values: linear/quadratic/product: 0.239, categorical: 0.250, threshold: 1.660, hinge: 0.500

Feature types used: hinge linear quadratic

responsecurves: true

jackknife: true

maximumiterations: 2000

'Equal Training Sensitivity and Specificity' and 'Equate Entropy of Thresholded and Original Distributions' thresholds and omission rates:

West model

ETSS-EETOD-Description

22.33-8.285-Cumulative threshold

0.345-0.156-Logistic threshold

0.150-0.259-Fractional predicted area

0.150-0.019-Training omission rate

East model

ETSS-EETOD-Description

44.108-10.106-Cumulative threshold

0.466-0.221-Logistic threshold

0.147-0.467-Fractional predicted area

0.147-0.000-Training omission rate