H.D.Eva E.E. de Miranda C.M. Di Bella V.Gond O.Huber M.Sgrenzaroli S.Jones A.Coutinho A.Dorado M.Guimarães C.Elvidge F.Achard A.S.Belward E.Bartholomé A.Baraldi G.De Grandi P.Vogt S.Fritz A.Hartley A VEGETATION MAP OF SOUTH AMERICA MAPA DA VEGETAÇÃO DA AMÉRICA DO SUL MAPA DE LA VEGETACIÓN DE AMÉRICA DEL SUR
20
Embed
H.D.Eva E.E. de Miranda C.M. Di Bella V.Gond O.Huber M.Sgrenzaroli S.Jones A.Coutinho A.Dorado M.Guimarães C.Elvidge F.Achard A.S.Belward E.Bartholomé.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
H.D.Eva E.E. de Miranda C.M. Di Bella V.Gond O.Huber M.Sgrenzaroli S.Jones A.Coutinho A.Dorado M.Guimarães C.Elvidge F.Achard A.S.Belward E.Bartholomé
A.Baraldi G.De Grandi P.Vogt S.Fritz A.Hartley
A VEGETATION MAP OF SOUTH AMERICAMAPA DA VEGETAÇÃO DA AMÉRICA DO SUL
MAPA DE LA VEGETACIÓN DE AMÉRICA DEL SUR
Contributing Institutions
Venezuela
Southern Cone
Amazon forest
Regional Experts working on data
Brazil
South America Map Production
• Multi-sensor approach- Humid forests detected using the ERS ATSR-2- Flooded forests ecosystems detected using the JERS-
1 RADAR- Urban areas selected using the DMSP ‘night lights’- Remaining land cover from SPOT VGT- Montane forests from G5 TOPO DEM (ammended)
Humid forests detected using the ERS ATSR-2
-Over 1000 images to create a mosaic based on highest surface temperature – “tropical dry season”
-Unsupervised spectral clustering
-Class labeling for humid forests and non-forests
ATSR-2
1 km resolution: 500 km swath: Green / Red / NIR / SWIR and TIR channels
ATSR-2 view of Rondonia
- R/G/B SWIR/NIR/Red
JERS-1 RADAR for flooded forests
- Two JERS-1 Mosaics – high water and low water
-Radar backscatter is increased by the the ‘double bounce’ off water and trees – high backscatter shows flooded forests
-The difference between the two images shows up seasonally flooded areas
Land cover from the SPOT VGT data
-Preparation of ‘seasonal’ mosaics from S10 data
-‘Winter’ ‘Spring’ ‘Summer’ ‘Autumn’ – selected on lowest SWIR (thresholded)
-Composited to the full year (Red NIR and SWIR)
-Humid forests (ATSR) mask
-Unsupervised classification 60 classes to remaining area
-Class labeling
-Extraction of particular areas from seasonal mosaics (e.g. removal of Snow)
Creation of seasonal mosaics from S10 product
Jan-Mar Apr-Jun July-Sept Oct-Dec
•Combining of VGT seasonal images•Masking of evergreen forest (use of ATSR forest)
•Unsupervised clustering to 60 classes
•Class labeling and aggregation with seasonal profiles