Tracking land cover change in the Motueka catchment using ENVISAT
David Pairman(1) , Stephen McNeill(1)
, Heather North(1)
, and Stella Belliss(1)
PO Box 69,
A two year series of 21 MERIS images has been used to asses the sensor’s ability to track land cover change in the Motueka Catchment, New Zealand. Ground truth is provided by Landsat TM imagery, aerial photography, forestry company records and field visits. The changes of interest relate to the modelling of runoff issues, where change or removal of the vegetation cover can increase the silt or nutrients reaching the Motueka River.
Significant vegetation changes, such as the clearing of forest stands or the ploughing of agricultural fields, are readily seen provided they are large enough in area. Progressive clear-felling of forest blocks can also be observed in time series analysis of the data.
However, typical changes over a monthly timescale are too small for a per-pixel classification approach. Our planned approach, using sub-pixel change detection methods, suffers from problems ortho-rectifying the full resolution level 2P images to the required accuracy. This is in part due to a nearest neighbour resampling step used in the generation of the level 2P data. An alternate approach is demonstrated where changes are first identified in higher spatial but lower temporal resolution TM data. The MERIS image series is then used to locate these changes more accurately in time.