We used Maxent (43) to calibrate ecological niche models for each species or lineage. In view of the difficult calibration challenges, we set the convergence threshold to 10−6. We generated 10 bootstrapped replicate models (50% subsampling) and extracted the median output across the 10 replicates for each. We then extracted the pixel-by-pixel median of the five replicate subsamplings, each representing the median of the 10 bootstrap replicates. Given emerging realization of contrasts between model information content and model transfer ability (44), we did not use Akaike Information Criterion (AIC)–based model selection approaches.

Once present-day and the two LGM projections of niche models were in hand, we established appropriate thresholds for separating prediction of presence versus absence on the basis of the least training presence threshold approach (45), but modified to take into account the expected error parameter E (46). That is, instead of setting a threshold at the value that includes 100% of the training occurrence data, we used the threshold that includes (100–E)% of the training occurrence data, which is thus a more restricted area but takes into account the possible presence of noise in the occurrence data. We explored E values of 0, 5, and 10% to provide a variety of views of confidence in model projections. Thresholds were established on the basis of the present-day models and then applied to LGM maps.

To test general ideas regarding the stability of potential distributional areas through time, we subtracted the present-day data layer from the LGM projection for each species. We averaged these difference maps across all species. After removing all ecoregions (41) not corresponding to AB biomes, we also removed all areas termed either varzea (seasonally flooded forest) or seasonal forests. We then used zonal statistics approaches (ArcGIS 10.3) to calculate changes in suitable areas within each of the interfluvia across the region.

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.