![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
Mapping the vegetation colonization on recent lava flows using spectral unmixing of moderate spatial resolution satellite images: Nyamuragira volcano, D. R. Congo |
VerfasserIn |
Long Li, Matthieu Kervyn, Frank Canters |
Konferenz |
EGU General Assembly 2014
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 16 (2014) |
Datensatznummer |
250089567
|
Publikation (Nr.) |
EGU/EGU2014-3773.pdf |
|
|
|
Zusammenfassung |
In volcanic areas, vegetation colonizes recently erupted lava flows and expands over time.
The fraction of vegetation is therefore likely to provide information on lava flows’ age.
Individual lava flows are usually not well resolved on satellite imagery due to the coarse
spatial resolution: one pixel on the imagery is a mixture of mainly lava and vegetation. In
order to solve the mixed pixel problem, many different methods have been proposed
among which linear spectral unmixing is the most widely-used. It assumes that
the reflectance of the mixed pixel is the sum of the reflectance of each pure end
members multiplied by their proportion in the pixel. It has been frequently used
in urban area studies, but no efforts have yet been made to apply it to volcanic
areas.
Here, we demonstrate the application of linear spectral unmixing for the lava flows
of Nyamuragira volcano, in the Virunga Volcanic province. Nyamuragira is an
active volcano, emitting over 30 lava flows in the last 100 years. The limited access
to the volcano due to social unrest in D. R. Congo justifies the value of remote
sensing techniques. This shield volcano is exposed to tropical climate and thus
vegetation colonizes lava flows rapidly. An EO-1 ALI image (Advanced land imager
mounted on Earth Observing -1 Satellite) acquired over Nyamuragira on January 3,
2012 at spatial resolution of 30 m was processed with minimum noise fraction
transform and end member extraction, and spectrally unmixed by linear mixture
modelling technique into two types of lava, and one or two types of vegetation.
The three end member model is better in terms of the RMSE and the expected
spatial distribution of end members. A 2 m resolution Pleiades image acquired
on January 21, 2013 and partly overlapping with the ALI image was taken as the
reference image for validation. It was first classified using a supervised pixel-based
classification technique and then compared to the proportion image derived from the
ALI image. Results show that accuracy depends on the size of moving window of
validation samples. We find a best fit (R2 >0.8) between the two datasets when
using a 180 x 180 m2validation sample. We also find that vegetation proportion
have a strong linear correlation with the normalized difference vegetation index
(NDVI). When applied to the entire ALI scene, the proportion of vegetation on
the recent flows is shown to be mostly controlled by the age of the lava surface
and the proximity to the flow boundary. This technique opens the perspective to
further characterize the dynamics of vegetation recovery on fresh volcanic surface. |
|
|
|
|
|