2017 ESA Annual Meeting (August 6 -- 11)

COS 35-7 - What are SDMs really about? Exploring the conceptual landscape of species distribution modeling literature using automated content analysis

Tuesday, August 8, 2017: 10:10 AM
D131, Oregon Convention Center

ABSTRACT WITHDRAWN

Emily B. McCallen1, Rod N. Williams2 and Songlin Fei2, (1)Forestry and Natural Resources, Purdue University, Lafayette, IN, (2)Forestry and Natural Resources, Purdue University, West Lafayette, IN
Emily B. McCallen, Purdue University; Rod N. Williams, Purdue University; Songlin Fei, Purdue University

Background/Question/Methods

Ecologists have always been interested in exploring the relationships between species and their habitat. Recent attempts to quantify the ecological niche have resulted in species distribution models (SDMs) which quantitatively link habitat variables to species occurrences. SDM predictions are often interpolated to produce potential range maps or extrapolated to distant geographical areas or hypothetical environmental conditions.

Available SDM literature has grown exponentially with methodological best practices changing rapidly. Traditional reviews are often limited and unable to cover the full scope or depth of a body of literature. Automated context analysis (ACA) is a tool that uses text-parsing and machine learning to identify and explore concepts within a body of text. This study uses ACA to explore changes in the conceptual landscape of SDM literature.

We used the program Leximancer to perform ACA on all available SDM abstracts. We searched for scientific articles and reviews on Scopus using a variety of terms related to habitat modeling. After downloading and exporting all abstracts, we manually checked all results to ensure they were topically appropriate and tagged them by type of manuscript (applied, methodological, or review). We explored general publication patterns over time as well as the conceptual landscape over the same period.

Results/Conclusions

We collected 5740 articles related to habitat modeling ranging from 1986-2016. We started our temporal analysis in 1994 since it was the first year with at least 10 available articles. We saw a steady increase in the number of available articles over time. While growth rate in the annual number of articles published has slowed in the past five years, an upward trend is still present with a maximum number of 649 articles published in 2016. Beginning in 1999, the proportion of applied articles has never fallen below 80%, indicating that SDMs are commonly used to answer ecological questions.

Results from the automated content analysis also revealed an application bias in the literature. The most commonly discovered concept in the literature was climate. This is likely related to the use of SDMs to predict species range shifts under climate change. Conservation and management were also commonly discovered concepts within the literature, reflecting the importance of the method within these fields. The only methodological concept seen within the top discovered concepts was variables, which reflects the importance of predictor variable choice within SDMs. Ultimately; we plan to use ACA to fully explore trends within the SDM literature over time.