Towards an automated satellite-based alarm system for global forest monitoring
(2010)
Files
These_SBontemps.pdf
Restricted Access - Adobe PDF
- 11.54 MB
Details
- Authors
- Supervisors
- Defourny, Pierre
- Abstract
- (en) Forests contribution to climate regulation and biodiversity conservation is widely recognized and their global monitoring has received particular attention. This thesis aims at developing an automated alarm system, based on high temporal resolution time series, which enables the near-real time detection of active forest change at sub-continental scales while accounting for the forest variability in space and over time. The concept of land surface change is first discussed in order to define the scope of the land surface monitoring systems. An automated and statistically-based forest change detection method is then developed. The method is based on a per-object approach applied in a multitemporal context and quantitatively integrates the forest variability in space and over time. Based on SPOT-VEGETATION time series, this method proves to be able to yearly detect deforestation hot-spots in the tropical forests of Brazil and Borneo with accuracies higher than 80%. Its efficiency is also demonstrated in the Mediterranean forest ecosystem using MERIS time series. In parallel, a multi-threshold analytical framework is implemented for the forest change detection and a decision-tool based on the ROC curves is implemented to select the appropriate threshold value according to the monitoring objectives. Accordingly, by taking advantage of high temporal resolution time series, the developed method demonstrates its efficiency to detect active forest change at subcontinental scales with reliability and on a yearly basis. These are valuable findings to move towards a regular and global alarm system for forest monitoring.
- Affiliations
Citations
Bontemps, S. (2010). Towards an automated satellite-based alarm system for global forest monitoring. https://hdl.handle.net/2078.5/260934