The impact of abrupt land cover changes by savanna fire on northern Australian climate: A grid computing approach

About 25% of Australia is covered by tropical savanna ecosystems. In these areas dry-season fires are the biggest natural and anthropogenic disturbance with 20% of this area being burnt or fire-affected on average during the last decade; this means that about 70% of the Australian extent of burning per year is caused by savanna fire. These fires result in meteorological impacts through the abrupt fire-induced land-cover changes as they modify the surface-atmosphere coupling. Results of initial sensitivity studies suggest not only a local response of the atmospheric system but also larger scale effects influencing northern Australian climate. This is especially vital as future climate change scenarios point to a more intense, severe fire regime.

To quantify the impacts we use the global climate model C-CAM in a stretched grid implementation with a 60 km resolution over Australia. The effects of fire and the temporal evolution of subsequent vegetation re-growth are characterized by a set of forcing perturbations (timing of fires, intensity, area burned, length of re-growth period) which are derived from identified ranges using relatively broad intervals. In order to have statistically stable results, improve the signal-to-noise ratio and cover a broader range of the natural system dynamics, 90 independent scenarios spanning the years from 1979 to 1999 are deployed. The change of vegetation properties is handled by a new scheme in the C-CAM land surface model, which modifies albedo, leaf area index, roughness length, and vegetation coverage on a grid-cell basis as a function of time and fire intensity.

A grid computing paradigm is the core component of the project. We apply the parametric modelling engine NIMROD/G (Buyya et al. 2000) for all aspects of process control, error, resource and data handling, as well as the distribution of the modelling system to cluster sites around the globe. Around 150 CPUs are made available for the overall experiment. The system is interactively set up and controlled via the NIMROD portal and viewer web-interfaces. Specific challenges discussed here include the long runtime of each experiment, the heterogeneous computing environments and the large data volume.

Buyya, R., Abramson, D. and Giddy, J., 2000, Nimrod/G: An Architecture of a Resource Management and Scheduling System in a Global Computational Grid, HPC Asia 2000, May 14-17, pp 283-289, Beijing, China.