Evaluation of vegetation-fire dynamics in the Okefenokee National Wildlife Refuge, Georgia, USA, with a Bayesian belief network

Vegetation response to wildfire has been studied extensively in upland ecosystems, but fire effects on temperate wetlands are less understood. We evaluated vegetation response to extensive wildfire in wetlands of Okefenokee National Wildlife Refuge (ONWR), USA, with a spatially explicit Bayesian belief network model informed with data recorded during 1990–2012. We assessed model accuracy and effects of fire frequency on vegetation composition with predictive scenarios of fire absence or a fire return interval (FRI) every 5 or 10 years during 2012–2032. In fire absence, shrubs increased 100%, primarily in the northern half of the Refuge, while the herbaceous class that was widespread in 2012 was eliminated. Areas dominated by forest during the past ~65 years were maintained with the 5- and 10-year FRI. Herbaceous-dominated areas maintained with the 5-year FRI decreased (90%) with the 10-year FRI. Shrub coverage increased with fire (17%, 5-year FRI; 20%, 10-year FRI), while scrub/shrub decreased (12%; 5-year FRI) or increased (6%; 10-year FRI). A 5-year FRI during conditions promoting severe fire may maintain the distribution of herbaceous and forested areas that followed an extensive drought and fires in 2011, and may limit scrub/shrub expansion that previously occurred with longer FRIs in the ONWR.