To elaborate our model we removed uncertain localities such as those with no coordinates and all records with coordinates laying down in sites with estimated elevations below 1,200 m and above 3,150 m.
The habitat suitability model generated in Maxent showed areas suitable in climatic terms for this species in Sierra Nevada de Santa Marta and very few pixels in serrania de San Lucas and sierra de La Macarena. These areas are not known to be occupied by the species and were excluded from the potential distribution map of this hummingbird.
Usually, it has been regarded that this species is not present in the pacific slope of the West Andes (Hilty & Brown 1986). Nevertheless, distribution data from EBIRD (2016) shows that the species is also present in that slope of the Andes.
Distribution of specimens, according to BioMap, suggests at least two possible areas of intergradation. One possible area (≈ 2,464 km2) of intergradation between subspecies caudatus and kingii, located between Norte de Santander and Santander, and another possible area (≈ 19,465 km2) of intergradation between subspecies mocoa and emmae, located between Huila, Cauca, Nariño and Putumayo.
Assuming that the distribution of the species may have filled the complete climatic model generated, its distribution today in remnants of forest is about 57,279 km2, which corresponds to a loss of 6 % of its potential original distribution due to deforestation.
Regularized training gain is 1.917, training AUC is 0.953, unregularized training gain is 2.054.
Algorithm converged after 980 iterations (37 seconds).
The follow settings were used during the run:
464 presence records used for training.
10462 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
'Equal Training Sensitivity and Specificity' and 'Equate Entropy of Thresholded and Original Distributions' thresholds and omission rates:
0.108-0.147-Fractional predicted area
0.108-0.026-Training omission rate