OOS 50-6 - Combining models and data to understand vegetation function across timescales

Friday, August 11, 2017: 9:50 AM
Portland Blrm 256, Oregon Convention Center
David J.P. Moore1, Andrew M. Fox2, Francesc Montane3, Amy R. Hudson4, Natasha MacBean4, M. Ross Alexander1, Mallory L. Barnes1, Ave Arellano4 and William K. Smith1, (1)School of Natural Resources and the Environment, University of Arizona, Tucson, AZ, (2)University of Arizona, AZ, (3)School of Natural Resources, University of Arizona, (4)University of Arizona

The rise in atmospheric CO2 directly alters the Earth’s climate and this has significant impacts on the livelihoods and well-being of citizens in the USA. The global land carbon (C) sink, calculated as the difference between human emissions and carbon accumulating in the atmosphere and in the oceans, has grown since 1960s, indicating that oceans and vegetation are absorbing anthropogenic C. Forest carbon uptake represents a significant portion of the land carbon sink and forest loss represents both a source of C emissions and the lost opportunity to take up and store C in natural systems. Understanding how forest carbon storage will change in future depends on the ecological and environmental controls of forest ecosystem processes and embedding this understanding in mathematical models. This poses two problems; first how do forest ecosystems work across scales and second how can we draw upon current observations of forest ecosystems to parameterize models and forecast carbon exchange over decades to centuries.

We use a data assimilation (DA) framework as both a diagnostic tool to understand ecological processes and a means to optimize mathematical models. DA is a general term for methods that systematically combine information from observations with information from a model to achieve an understanding of the system that is more accurate than the observations or the model independently. DA provides a statistical framework to use observations to estimate model states and parameters, evaluate alternative model structures, and quantify uncertainties in model predictions.


Using the SIPNET model (Simplified PhotosyNthesis and EvapoTranspiration; a lumped parameter ecosystem model), we have found that Eddy covariance observations can provide information on fast processes such as photosynthesis and total ecosystem respiration and demonstrate that the seasonal covariance of carbon balance and evapotranspiration (ET) can be exploited to provide information on the partitioning of ET into evaporation and transpiration. In addition, model structural modifications can be made to test high level hypotheses about how seasonal controls of respiration function and about the role of roots, microbes and C substrate availability in below-ground processes. To infer regional and global controls over vegetation function we have adopted a formal Land Surface Model; the Community Land Model. Initial findings at larger geographic and longer time scales are promising but highlight some ecological limitations of Land Surface Models. We discuss how tree carbon allocation, tree mortality and changes in forest communities might be studied using these approaches.