The Effects of Sample Size on Ecological Niche Models: Analysis Using the California Ground Squirrel
Paula Ceballos
Department of Biology
Faculty Supervisor: Robert Boria
Ecological Niche Models (ENMs) help us identify potential suitable conditions for species of interest through time and space. They do so by comparing the overall environmental conditions available to the species with the conditions at localities where the species occurs. Identifying the number of occurrence localities needed to model a species potential distribution is often overlooked. In this study, we model the effects of varying sample size of the California ground squirrel (Otospermophilus beechyi). First, we obtained locality data from GBIF (1,066 unique records) and applied a spatial filter to reduce sampling biases. We generated random localities for differing sample sizes, 10, 25, 50, 75, 100, 250, and 500 (10 replicates each). We used evaluation statistics to identify our model's discriminatory ability (AUC) and omission rates. Generally, performance statistics converged despite significant differences in dataset sizes. We were able to generate very similar ENMs with only 20% of the total data. Using a large number of occurrence records may not be necessary for ENMs, and in fact may hinder model performance. Future work includes comparing our findings with a different species, the Merriam’s Chipmunk (Tamias merriami), to compare results between a species with a broad geographic distribution and another with a more limited spatial range.