Applicability of CSD-based resilience analyses to remotely sensed Vegetation Indices in the Tropics

Tropical forests are vital for climate change mitigation as carbon sinks. Yet, research suggests that climate change, deforestation and other human influences threaten these systems, potentially pushing them across a tipping point where the tropical vegetation might collapse into a low-treecover state. Signs for this trend are reductions of resilience defined as the system's capability to recover from perturbations. When resilience decreases, according to dynamic system theory, a critical slowing down (CSD) induces changes in statistical measures such as the variance and the autocorrelation. This allows to indirectly examine resilience changes in the absence of observations of strong perturbations. Yet, deriving estimates of the statistical measures indicating resilience changes based on CSD impose several assumptions on the system under observation. For tropical vegetation, it is not obvious that these assumptions are fulfilled.

Additionally, the conditions of tropical rainforests pose difficulties on the observation of the vegetation. Among other factors, cloud cover, aerosols, and the dense vegetation hinder the reliable retrieval of Vegetation Indices (Vis), especially from data gathered in the optical spectrum. Thus, such data might not be suitable for resilience analyses based on CSD, even if the theory is applicable in principle.

We investigate the different assumptions of CSD and test them on a diverse set of remotely sensed VIs. Hereby, we establish a framework that allows to decide whether a specific dataset is appropriate for resilience analyses based on CSD.