Modelling Ecosystem Processes

We use models to encapsulate our understanding of natural systems. Models are useful because they allow us to test how well we understand nature, they allow us to fill gaps and to process data into more useful forms, and they also allow us to make predictions in time or projections in space. In the Biosphere Programme our modelling work is intricately linked to experimental and observational science. We believe that a close link between data and models fosters good science, and that the labelling of scientists as either "modellers" or "field workers" is artificial and redundant.

Key people: Mathew Williams, Tim Hill, Luke Spadavecchia, Dan Metcalfe, Rui Zhang, John Moncrieff, Rosie Fisher, Lisa Wingate.

Several models developed within the programme are available for use by the wider community - follow the Software Tools link to the left

Current research directions: data assimilation in ecosystem modelling

Ecologists have begun to generate extensive measurements of system biogeochemistry, but analysis of these time series data is highly complex. Measurements are never comprehensive, and critical components of the system are difficult or impossible to measure (e.g. root respiration and other below ground processes). We have begun to use data assimilation techniques to advance analyses. We have recently shown how time series of various data (sap flow, soil respiration, tree diameters, root biomass) can be combined with a mass balance model of C flows, via a technique called data assimilation, to produce a consistent analysis of ecosystem C cycling over annual time scales. The model provides a means to propagate information from well measured state variables (e.g. time series of leaf area index, soil respiration) to other poorly measured ones (e.g. root biomass), via constraints from the mass balance modelling. Data assimilation is also a powerful means of model testing – multiple model structures can be compared to find which is most consistent with the multiple data sets. Our Data Assimilation Linked Ecosystem Carbon (DALEC) model uses a very simple box model (nine parameters govern the movement of C among five pools), which makes model testing simple and transparent.

Williams, M., P.A. Schwarz, B.E. Law, J. Irvine & M. Kurpius (2005). An improved analysis of forest carbon dynamics using data assimilation. Global Change Biology 11, 89-105.

We have five current projects where modelling forms a core focus.

1. Carbon cycling in arctic terrestrial ecosystems. In collaboration with Gus Shaver (MBL, Woods Hole) and Mark Van Wijk (University of Wageningen, Netherlands), funded by NSF. For more details, click here
2. Carbon exchange in complex terrains. In collaboration with Bev Law and Larry Mahrt (Oregon State University), funded by NASA.
3. Tropical forests and global change. In collaboration with Yadvinder Malhi, Patrick Meir and John Grace (University of Edinburgh), funded by NERC and EU.
4. Carbon dynamics of miombo woodlands of Mozambique. In collaboration with John Grace (Edinburgh), EU funded project.
5. Data assimilation and ecosystem modelling. In collaboration with NERC Centre for Terrestrial Carbon Dynamics.

See also Mathew Williams's homepage