Understanding the effect of fire on vegetation composition and gross primary production in a semi-arid shrubland ecosystem using the Ecosystem Demography (EDv2.2) model

摘要

Wildfires in sagebrush ( Artemisia spp.)-dominated semi-arid ecosystems in the western United States have increased dramatically in frequency and severity in the last few decades. Severe wildfires often lead to the loss of native sagebrush communities and change the biogeochemical conditions which make it difficult for sagebrush to regenerate. Invasion of cheatgrass ( Bromus tectorum ) accentuates the problem by making the ecosystem more susceptible to frequent burns. Managers have implemented several techniques to cope with the cheatgrass–fire cycle, ranging from controlling undesirable fire effects by removing fuel loads either mechanically or via prescribed burns to seeding the fire-affected areas with shrubs and native perennial forbs. There have been a number of studies at local scales to understand the direct impacts of wildfire on vegetation; however there is a larger gap in understanding these impacts at broad spatial and temporal scales. This need highlights the importance of dynamic global vegetation models (DGVMs) and remote sensing. In this study, we explored the influence of fire on vegetation composition and gross primary production (GPP) in the sagebrush ecosystem using the Ecosystem Demography (EDv2.2) model, a dynamic global vegetation model. We selected the Reynolds Creek Experimental Watershed (RCEW) to run our simulation study, an intensively monitored sagebrush-dominated ecosystem in the northern Great Basin. We ran point-based simulations at four existing flux tower sites in the study area for a total of 150?years after turning on the fire module in the 25th?year. Results suggest dominance of shrubs in a non-fire scenario; however under the fire scenario we observed contrasting phases of high and low shrub density and C 3 grass growth. Regional model simulations showed a gradual decline in GPP for fire-introduced areas through the initial couple of years instead of killing all the vegetation in the affected area in the first year itself. We also compared the results from EDv2.2 with satellite-derived GPP estimates for the areas in the RCEW burned by a wildfire in 2015 (Soda Fire). We observed moderate pixel-level correlations between maps of post-fire recovery EDv2.2 GPP and MODIS-derived GPP. This study contributes to understanding the application of ecosystem models to investigate temporal dynamics of vegetation under alternative fire regimes and post-fire ecosystem restoration.