Topic Modeling of the Spatiotemporal Interlinkages of Socio-Environmental Challenges Facing the Mississippi River Basin

Keywords: Mississippi River Basin; Machine Learning; Topic Modeling; Socio-Environmental Challenges; River Management; Climate Change
Abstract Type: Poster Abstract

Abstract

The human-environmental challenges facing the Mississippi River Basin (MRB) exist on range of spatial and temporal scales, ranging from deltaic hypoxia due to agricultural runoff, to urban sprawl onto floodplains, to changes in migratory bird behavior over headwaters. These challenges may be unique or interlinked among individual states, sub-basins, or points of time in prior MRB research, an interdisciplinary awareness of which is essential in order to determine research priorities that build upon existing human-environmental knowledge. To that end, we present in this research a topic modeling approach that synthesizes existing research about the MRB published from 1990 to 2023 and consequently identify common topics across publications and government reports on a range of spatial (state, sub-basin, basin-wide) and temporal (annual, decadal) timescales. The specific approach is one of Latent Dirichlet Allocation, a machine learning algorithm that probabilistically quantifies the likelihood of words and phrases occurring together across documents. Preliminary findings suggest that research priorities across the MRB have evolved over time, progressing from site-specific to basin-wide considerations and an increased valuation of climate change and stakeholder/partner engagement. Topics such as invasive fish species cataloging and nutrient overloading of rivers have remained common since 1990 and across the majority of states comprising the MRB's spatial domain. Findings from this topic modeling approach will be useful for identifying priority directions for future MRB research, as well as guide decisions for funding future research into the MRB's growing need for climate adaptation management.