Heliomaster longirostris

NOTES

To elaborate our models we removed uncertain localities such as 'Bogota skins', those localities with no coordinates and all records with coordinates laying down in sites with estimated elevations above 2,164 m.

Initially we produced a unique model with all data. Nevertheless, prediction of presence east of the Andes was very poor. Thus we split the data in two sets and modelled separately the range west and east of the Andes, which improved predictions. In each case the habitat suitability model generated in Maxent showed to some extent suitable areas in the side not modelled. Areas that did not correspond to the side modelled were deleted in each case from the potential distribution maps of this hummingbird. Possibly the species is further distributed north in 'Los Llanos' de Arauca. However, the are no known records pointing to that and areas were not suggested as possibly suitable.

Assuming that the distribution of the species may have filled the complete climatic model generated for the areas west of the Andes, its distribution today in remnants of forest is about 73,291 km2, which corresponds to a loss of 67 % of its potential original distribution due to deforestation. Nonetheless, this species favours edges, secondary vegetation and plantations and therefore possibly deforestation has not negatively affected greatly its populations. On the other hand, assuming that the distribution of the species may have filled the complete climatic model generated for the areas east of the Andes in the Amazon region, its distribution today in remnants of forest is about 474,504 km2, which corresponds to a loss of 9 % of its potential original distribution due to deforestation in that region.

MODEL METADATA

West of the Andes model

Regularized training gain is 1.707, training AUC is 0.955, unregularized training gain is 2.098.

Algorithm terminated after 2000 iterations (75 seconds).

The follow settings were used during the run:

188 presence records used for training.

10183 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 of the Andes model

Regularized training gain is 1.389, training AUC is 0.911, unregularized training gain is 1.955.

Algorithm converged after 1060 iterations (38 seconds).

The follow settings were used during the run:

38 presence records used for training.

10038 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: 0.227, categorical: 0.250, threshold: 1.620, 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 of the Andes model

ETSS-EETOD-Description

21.929-9.604-Cumulative threshold

0.320-0.153-Logistic threshold

0.106-0.181-Fractional predicted area

0.106-0.021-Training omission rate

East of the Andes model

FCV5-ETSS-EETOD-Description

5.000-24.008-20.784-Cumulative threshold

0.080-0.183-0.161-Logistic threshold

0.552-0.211-0.249-Fractional predicted area

0.026-0.211-0.105-Training omission rate