|
Titel |
First Results of the Performance of the Global Forest/Non-Forest Map derived from TanDEM-X Interferometric Data |
VerfasserIn |
Carolina Gonzalez, Paola Rizzoli, Michele Martone, Christopher Wecklich, Jose Luis Bueso Bello, Gerhard Krieger, Manfred Zink |
Konferenz |
EGU General Assembly 2017
|
Medientyp |
Artikel
|
Sprache |
en
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 19 (2017) |
Datensatznummer |
250143638
|
Publikation (Nr.) |
EGU/EGU2017-7380.pdf |
|
|
|
Zusammenfassung |
The globally acquired interferometric synthetic aperture radar (SAR) data set, used
for the recently completed primary goal of the TanDEM-X mission, enables a big
opportunity for scientific geo-applications. Of great importance for land characterization,
classification, and monitoring is that the data set is globally acquired without gaps and
includes multiple acquisitions of every region, with comparable parameters. One of the
most valuable maps that can be derived from interferometric SAR data for land
classification describes the presence/absence of vegetation. In particular, here we
report about the deployment of the Global Forest/Non-Forest Map, derived from
TanDEM-X interferometric SAR quick-look data, at a ground resolution of 50 m by
50 m.
Presence of structures and in particular vegetation produces multiple scattering
known as volume decorrelation. Its contribution can be directly estimated from the
assessment of coherence loss in the interferometric bistatic pair, by compensating for all
other decorrelation sources, such as poor signal-to-noise ratio or quantization noise.
Three different forest types have been characterized based on the estimated volume
decorrelation: tropical, temperate, and boreal forest. This characterization was then used in a
fuzzy clustering approach for the discrimination of vegetated areas on a global
scale.
Water and cities are filtered out from the generated maps in order to distinguish volume
decorrelation from other decorrelation sources.
The validation and performance comparison of the delivered product is also presented,
and represents a fundamental tool for optimizing the whole algorithm at all different
stages.
Furtheremore, as the time interval of the acquisitions is almost 4 years, change detection
can be performed as well and examples of deforestation are also going to be included in the
final paper. |
|
|
|
|
|