Species distribution modeling of seagrass in the Gulf of Mexico

Keywords: seagrass, Gulf of Mexico, species distribution modelling, random forest, machine learning
Session Type: Virtual Poster Abstract
Day: Friday
Session Start / End Time: 2/25/2022 08:00 AM (Eastern Time (US & Canada)) - 2/25/2022 09:20 AM (Eastern Time (US & Canada))
Room: Virtual 38

Abstract

Seagrass meadows function as underwater forest ecosystems integral to biological, chemical, and physical processes. Species distribution modeling is useful in characterizing the dimensions of environmental conditions associated with observed species distribution and predicting spatial distribution of the species in regions of interest. Seagrass habitats in the northern Gulf of Mexico (GmM) have experienced long-term, progressive degradation/decline and increasing vulnerability resulting from prior and ongoing anthropogenic disturbances, as well as environmental stressors. We prepared and integrated environmental variable and seagrass presence/absence datasets, as well as predict the seagrass distribution in the northern GoMusing random forest machine learning. Environmental variables include dissolved oxygen, nitrate, phosphorous, salinity, silicate, and temperature that were derived from the global Ecological Marine Unit (EMU) 3D point mesh dataset, SRTM (Shuttle Radar Topography Mission) seafloor geomorphology data, surficial sediment median grain size, and median bottom shear stress estimates (BSS). Of the incorporated environmental variables, RF model identified median sediment grain size, median] BSS, silicate, nitrate, and phosphate as being the most important predictors. Important variables identified and the predicted seagrass spatial distribution map give insight into the environmental drivers and suitable environmental range of seagrass distribution spanning the northern Gulf of Mexico shelf regions, which could extend supplemental function in efforts to preserve seagrass habitats.