Fusing data and models to forecast disturbance impacts on ecosystems: past, present, and future

Abstract:

Disturbance plays a major role in shaping ecosystems, but their occurrence is stochastic and their representation in ecosystem models is widely considered to be inadequate. Both these factors place significant limits on our ability to forecast ecosystems in a changing world. Herein we address three major challenges in disturbance modeling, focusing on recent advances and emerging opportunities. First, we assess the ability of models to capture emergent observational patterns and experimental manipulations using a range of data types and across a wide range of time scales, from sub-annual (eddy flux) to interannual (inventory, tree rings) to centennial (fossil pollen and charcoal, settlement era surveys). Second, we present advances in how models can represent specific disturbance processes (insects and pathogens, non-harvest forest management) and the scaling of disturbance processes in regional- to global-scale models. Third, we address the question of how to assimilate multiple sources of disturbance data across multiple spatial and temporal scales to constrain the pools and fluxes within models. In particular, because disturbances are stochastic we want to update model projections based on field and remotely-sensed data. However, rather than driving models with a particular external data that is taken as ‘truth’, we explore examples of how to update models using multiple data sources while accounting for uncertainties in both the model and data. Finally, we discuss the implications of these advances and the opportunities they present for near-term carbon monitoring and long-term carbon cycle projections.