Estimating continental scaling of ecological driver-response feedbacks through model-data fusion

Background/Question/Methods

Rapid biological change is occurring around the world as a result of human development. Ecosystems are increasingly stressed by climate change, human land use decisions, atmospheric pollution and invasive species. In response to these drivers the biogeochemistry, ecohydrology and biodiversity of ecosystems is changing. NEON has been designed to provide the ability to observe both the causes of, and responses to, such change, producing data that can be used to understand, forecast and ultimately manage the changing biosphere. A challenge facing the use of these data is how to best make inference at regional and continental scales from ecological information available a limited number of study sites.

One approach is to use Earth System Models (ESMs) that simulate the carbon and water cycles through a process-based representation of the interactions and feedbacks between ecological drivers and responses. ESMs provide a framework in which NEON-derived terrestrial biological measurements, surface-atmosphere flux observations, landscape-scale airborne data and existing national datasets can be fused. We have done this in a rigorous statistical manner using data assimilation (DA) to produce an optimal analysis describing current ecosystem states at a full range of spatial scales. 

Results/Conclusions

We have developed an ensemble DA system for the Community Land Model (CLM). This is a state-of-art land surface model used in the fully coupled Community Earth System Model (CESM). Using this new system we have been able to constrain CLM with multiple data types and have developed a range of forward operators that can link CLM state and diagnostic variables to observations, taking in to account sub-model grid cell variability in plant functional types.

Here we show how tower data describing variation in carbon and water fluxes in response to climate drivers informs likely CLM carbon stocks and fluxes, both at tower sites and across a regional domain. We then modified this response to take into account past disturbance drivers. Whilst it is not possible to know accurately vegetation disturbance history at fine spatial scales, disturbance is reflected in present day spatial variability in satellite-derived leaf area index and biomass data. We used this information to adjust modeled carbon stocks and fluxes away from a simulated, undisturbed state to better describe biogeochemistry and ecohydrology at the continental scale and investigated the impact this has on future responses to climate and disturbance drivers.