A comparison of MODIS time series decomposition methods for estimating evergreen foliage cover

Gill, Tony, Armston, John, Phinn, Stuart and Pailthorpe, Bernard (2006). A comparison of MODIS time series decomposition methods for estimating evergreen foliage cover. In: Proceedings of the 13th Australasian Remote Sensing and Photogrammetry Conference (13 ARSPC). 13th Australasian Remote Sensing and Photogrammetry Conference, Canberra, (). 20-24 November 2006.


Author Gill, Tony
Armston, John
Phinn, Stuart
Pailthorpe, Bernard
Title of paper A comparison of MODIS time series decomposition methods for estimating evergreen foliage cover
Conference name 13th Australasian Remote Sensing and Photogrammetry Conference
Conference location Canberra
Conference dates 20-24 November 2006
Proceedings title Proceedings of the 13th Australasian Remote Sensing and Photogrammetry Conference (13 ARSPC)
Place of Publication Canberra
Publisher Spatial Sciences Institute
Publication Year 2006
Sub-type Fully published paper
ISBN 9780958136655
0958136653
Total pages not found
Collection year 2006
Formatted Abstract/Summary
In Australia, methods for decomposing time series AVHRR NDVI imagery into evergreen and raingreen signals for estimating evergreen and raingreen foliage cover have been developed. The objective of this study was to compare the accuracy of two complex time series decomposition models with a third, simpler, model for estimating evergreen foliage cover. This involved developing a validation approach and also determining which of the two MODIS vegetation indices (NDVI or EVI) was a better predictor of evergreen foliage cover.

MODIS vegetation index imagery was acquired over Queensland for the period from April 2000 to March 2006. The imagery was processed using each of the three algorithms to derive evergreen and seasonal estimates of NDVI and EVI. A relationship to extract estimates of evergreen vegetation cover from the evergreen VI imagery was developed by matching image pixels to woody foliage projective cover (FPC) data derived from airborne Lidar acquired at 19 sites in Queensland between 2003 and 2004. Validation of the results was undertaken via comparison with woody FPC estimates derived from independent Lidar data, woody FPC estimates derived from Statewide Landcover and Trees Study (SLATS) Landsat imagery and MODIS global continuous fields product (MOD44B).

Initial results indicate 80% correspondence between evergreen vegetation estimates derived from MODIS NDVI and Lidar data. The EVI estimates produce results significantly lower than NDVI. There is little difference in accuracy between the time series models. Early comparisons with the SLATS data indicate a reasonable correspondence of MODIS evergreen FPC to woody FPC. Where the MODIS evergreen FPC is significantly lower than the Landsat estimates correlates to areas of high raingreen FPC. These results highlight the potential application of time series decomposition for augmenting the discrimination of raingreen and evergreen FPC from Landsat imagery.
Subjects EX
291003 Photogrammetry and Remote Sensing
770406 Integrated (ecosystem) assessment and management
Q-Index Code EX

 
Versions
Version Filter Type
Citation counts: Google Scholar Search Google Scholar
Created: Thu, 23 Aug 2007, 22:22:18 EST