Untitled ItemA core model for forecasting the dynamics of a fire-prone ecosystem
Domaine:
environment and energygeospatialType de record:
model
Créateur:
Éditeur:
fig
Hôte:
Presentation given at the virtual International Statistical Ecology Conference (Sydney, Australia) June 22-26, 2020:
Abstract:Longitudinal ecological studies are commonly limited in their spatial grain and extent, relying on regular, continuous global measurements of vegetation properties from satellites to extrapolate to a broader context. This is readily feasible for ecosystems with easily interpretable changes in the temporal vegetation signal, but is far more complex for ecosystems subject to stochastic natural disturbance events and post-disturbance recovery trajectories. We have developed a time-series model to monitor and predict natural land surface processes and evaluate near real-time changes in the state of a crown fire-dependent and seasonally fluctuating ecosystem, the Fynbos of South Africa. The model allows near-real time estimation of above ground biomass, impacts on streamflow, and detection of anomalous vegetation signals such as high mortality, vegetation clearing, or invasion by alien trees. The modelling framework is particularly novel in that it also allows estimation of the factors that drive variation in post-fire recovery, such as climatic, edaphic, or vegetative properties such as plant traits. Our core model is highly flexible and ideal for exploring the drivers of multiple ecosystem functions within one framework. It also tests our understanding of the ecosystem, guiding further sampling and research, and iteritive updating of the model, data and forecasts.
Abstract:Longitudinal ecological studies are commonly limited in their spatial grain and extent, relying on regular, continuous global measurements of vegetation properties from satellites to extrapolate to a broader context. This is readily feasible for ecosystems with easily interpretable changes in the temporal vegetation signal, but is far more complex for ecosystems subject to stochastic natural disturbance events and post-disturbance recovery trajectories. We have developed a time-series model to monitor and predict natural land surface processes and evaluate near real-time changes in the state of a crown fire-dependent and seasonally fluctuating ecosystem, the Fynbos of South Africa. The model allows near-real time estimation of above ground biomass, impacts on streamflow, and detection of anomalous vegetation signals such as high mortality, vegetation clearing, or invasion by alien trees. The modelling framework is particularly novel in that it also allows estimation of the factors that drive variation in post-fire recovery, such as climatic, edaphic, or vegetative properties such as plant traits. Our core model is highly flexible and ideal for exploring the drivers of multiple ecosystem functions within one framework. It also tests our understanding of the ecosystem, guiding further sampling and research, and iteritive updating of the model, data and forecasts.