The GLOBCARBON Initiative: Massive Multi-Sensor, Multi-Year, Multi-Product Processing
The GLOBCARBON initiative developed a service generating fully calibrated estimates of at land products (multi-product), being Burnt Area Estimates and Global Biophysical parameters (LAI, fAPAR and VGCP). These products helps us understanding the spatial and temporal variation in carbon fluxes, which is essential to constrain models that predict climate change. They are generated quasi-independently of the original Earth Observation sources (multi-sensor). It is a clear ambition of the project to demonstrate the synergistic use of the data from multiple instruments for operational applications. This project proofed that this synergistic use can increase the confidence of satellite observations and can help in the creation of reliable time series of added-value products. The data is provided by both instruments on ESA (MERIS, AATSR and ATSR2) and non-ESA (VEGETATION 1 and 2) Earth Observation satellites. The project generates products for ten consecutive years (from 1998 up to 2008), whereas different combinations of data-sources are used upon availability (multi-year).
Processing such a massive amount of data has many implications for the operational environment.
This implies on the hardware level the setup of a adapted computer facility and on the software level an implementation where very much care goes to the performance of the developed algorithms. Different performance bottlenecks were identified and solved. The estimated processing time is 10 months, i.e. 1 year of data is processed in 1 month.
This processing system is using large archives of raw input data which are located in different geographical locations (MERIS and AATSR data in Italy, VEGETATION data in Belgium). When processing the entire processing chain at one location, the need to move a large amount of data from Italy to Belgium (or vice versa) created a logistical problem.
A solution which is envisaged to overcome this logistical problem is GRID computing. The purpose of a grid-compatible globcarbon processing system, is to bring the processing power and knowledge to the data (whether it is located in Italy or Belgium) instead of having to move large amounts of data to the processing site.
This paper describes the synergistic use of multiple sensor data and features discussion of the operational solutions to the problems of handling terabytes of coarse resolution Earth observation data and converting them into added value products.