The current operational method of land cover (use) change detection, as also reflected in current line of Copernicus land services, are based on bi-temporal analysis for two points in time. These can be performed at decision level (supervised approach) or at data level (unsupervised approach). The most common methods include image differencing/rationing, linear transforms, change vector analysis and image regression or visual interpretation. There is no consensus on the type of change detection algorithm that performs best. Different studies and comparisons report different results. Principally there is not one universal method that would be applicable within all land cover domains. Moreover, these methods rely on the assumption that pixels associated with land-cover changes present spectral values that are significantly different from pixels associated with unchanged areas. However, vegetated terrestrial ecosystems exhibit high temporal (seasonal) variation that is captured in the acquired bi-temporal images, which considerably weaken the general assumption of the methods.
The temporal trajectory approach assumes continuous seasonal development of vegetation. With this approach seasonal development curves for different seasons are compared to detect changes. The use of temporal trajectory analysis for the problem of change detection is a way of avoiding the problem of selecting optimal points in time for images to be compared. This is generally performed for medium-resolution imagery as these can provide the necessary temporal resolution. However, with availability of the high resolution image archives (as Landsat 5/7/8), it is possible to retrospectively monitor land ecosystem dynamics, while in future it can be continued by Sentinel-2 mission.
The TED-processor being developed consists of the three subsystems that include image pre-processing to artefact-free time series, local-adaptive dynamic land surface model and change detection decision support system. The performance of the processor is demonstrated on the case study focused on deforestation monitoring in the region of Sumava national park within the time period from 1984 to 2014.
The TED monitoring project is funded by ESA within the Czech Industry Incentive Scheme.
Gisat provides wide range of geoinformation services based on Earth Observation technology. It focuses on operational application of satellite mapping to monitor various aspects of our environment and development of dedicated web based platforms for geoinformation analysis and assessment
Web // E-mail // Tel:+420 271741935 // Fax: +420 271741936