Species distribution modeling for honeycreepers on Kauaʻi, Hawaiʻi informs conservation action

Published: 8 January 2026| Version 1 | DOI: 10.17632/6w34spkfhh.1
Contributors:
Jackson Alexander,
,
, Roy Gilb, Lisa Crampton

Description

We analyzed habitat suitability for five Hawaiian honeycreepers on Kaua'i, Hawai'i to generate life history and habitat use knowledge and to identify priority sites for conservation action. As our study species are vulnerable, threatened, and endangered species, we have removed all location data from the data published here. We instead include as input to Maximum Entropy (MaxEnt) a csv file with the extracted environmental variables at each presence point. MaxEnt can be run from just this table, though that is not how we ran it. Aside from specific locations, the attached data is the entirety of datasets we used in our analysis and all code needed to run the analysis. Any missing code represents a non-code based portion of our analysis. For example, we ran MaxEnt in the MaxEnt GUI, so there is no code to create our MaxEnt models. Full datasets with precise location data, if required, can be requested from Lisa "Cali" Crampton at cali@kauaiforestbirds.org.

Files

Steps to reproduce

First, we created 100 m buffers around each detection (anonymized here) using the script 'search_window'. Then, we calculated habitat variables within each buffer (see columns in files called ʻOccurenceBuffers100withLiDARʻ), generated background buffers and calculated habitat variables (files called ʻXXXX###_BackgroundBuffers100withLiDARʻ), then ran statistical tests (statistical_analysis.rmd) to compare the detections against landscape values. We used our detection points as input to MaxEnt in the MaxEnt GUI with the rasters in "Environmental Predictors (.asc)" as input predictors. Since we anonymized our detections, this step can be reproduced using MaxEnt or an R package with MaxEnt modeling capabilities, using the table 'ForestBirdDetectionsForMaxEnt22-23wLiDAR'. Then, we created range estimates using the output of MaxEnt for each species and the script 'maxent_threshold_MTP'. We evaluated the performance of our MaxEnt model using testing data from 2024, 'ForestBirdDetectionsForMaxEnt240101-20240716.csv'. Some other steps in our analysis, like summing the outputs of each species' MaxEnt model, were done in ArcGIS Pro and could not be reproduced in code. These are straightforward steps, though, that could be reproduced in any GIS.

Categories

Conservation Biology, Wildlife Distribution, Spatial Ecology, Environmental Niche Modeling

Funders

Licence