We apply content analysis on government documents containing ecological information relevant to a significant ecological disturbance - mountain pine beetle (MPB) outbreaks in the United States. The intent is to demonstrate a semi -automated approach that applies topic modeling to investigate policy responses to ecological disturbances, using latent Dirichlet allocation (LDA) and hierarchical Dirichlet process (HDP), and term frequency inverse document frequency (tf-idf) analysis. Results demonstrate how analysts and researchers are better able to understand what topics and focus areas government officials consider in relation to MPB disturbances. In the case study demonstrating the method's utility, documents found from before 1960 and until recent years demonstrate focus on outbreak area, tree mortality, research and services, management, infestation, outbreak control, fire, insect control, outbreak factors, and tree population. Terms such as fire, mortality, treatment, and outbreak reflect more recent U.S. government focus on MPB, while disease and infestation have become less of a focus in recent years. There are also varying differences and interests between how different parts (i.e., federal agencies versus congress) of the U.S. government focus on MPB, where mostly interests and focus are not aligned or do not match temporally. As a term, temperature has become a greater recent government focus, but there is general avoidance of the term climate change. The methods applied demonstrate the utility of topic modeling and tf-idf for understanding discourse and content in policy related to ecological disturbances. The tool created in this effort is provided freely as a way for scientists and researchers to extend its utility in ecological policy research.
1.UCL, Inst Archaeol, London, England 2.Univ Victoria, Dept Geog, Victoria, BC, Canada 3.Univ Georgia, Savannah River Ecol Lab, Warnell Sch Forestry & Nat Resources, Athens, GA 30602 USA
Recommended Citation:
Altaweel, Mark,Bone, Christopher,Abrams, Jesse. Documents as data: A content analysis and topic modeling approach for analyzing responses to ecological disturbances[J]. ECOLOGICAL INFORMATICS,2019-01-01,51:82-95