Research project S0/00/052 (Research action S0)
Context and objectives
The recent evolution of earth observing systems makes available to the users an increasing number of continuous temporal series acquired at different spatial and spectral resolutions. In order to deal with large continuous temporal series, automatic and global processing methods are required. In particular these must concern the capacity of producing useable cloud-free images at regular interval, i.e. the compositing procedure and the capability of detecting any significant signal change out of a large volume of data (VEGETATION sensor). Both issues are specifically tackled in the framework of this research project.
Methodology
• The mean compositing method (Vancutsem et al, 2005) was extended to global scale for the compositing step. The quality control required for this step is based on a collection of 110.000 pixels characterised, which allowed the identification of suited thresholds as well as the implementation of a statistical approach based on a discriminant analysis.
• The change detection methods implemented in this project is based on the detection of the temporal behaviour variations of the signal. The originality of the approach lies in the analysis of the temporal signal in all wavelengths and not only the use of a vegetation index as it is commonly carried out. This rigorous approach was made possible thanks to the spatial and temporal stability of the signal provided by the compositing method. This multispectral approach allowed refining the quality of the detection.
Results
The research activities lead to the production of several algorithms ready to be used by remote sensing specialists or by an automatic monitoring system in operational processing environment on a near real-time basis at the worldwide scale.
A processing chain intended to produce composites based on the mean compositing strategy was developed. This chain is very flexible and allows the user to optimise the compositing parameters according to its application and the regional conditions. The syntheses present a large spatial and temporal consistency in all wavelengths and their production is so fast that the whole archive can be reprocessed at any time.
An operational processing chain intended to monitor the conditions favourable to the development of the Locust was developed. This process, currently tested at the FAO, ensures the monitoring in real time and improves the quality of the detections.
An operational processing chain intended to monitor water bodies was developed. It allows the characterisation and the monitoring of temporarily and permanent water bodies as well as flooding events.
Tools intended to put the detection results at the disposal of the users were developed in the framework of the early warning application.
Products and services
Many scientific collaborations were developed with international institutions (FAO, NASA, ICRISAT, CIRAD Guyane, universities). In the framework of these collaborations, several products were computed:
- Global temporal series at different frequency over the years 2000 to 2005
- Decades detection of areas favourable to the development of the Locust over the years 2000 to 2005
- Detection and characterisation of water bodies over the north-Africa area (validated) and application of the methodology to the global scale (not yet validated) for years 2003 to 2005
- A global mask of coastal areas intended to the VEGETATION products users
