Hello there,
I have a question -
I am trying to find out the predictive ability of broadband vegetation indices (satellite NDVI, NDWI, ATSAVI) to estimate three comparable narrowband vegetation indices (ASD spectroradiometer NDVI, NDWI, ATSAVI) or leaf area index (LAI) using 3 satellite images [SPOT-5 (10 m), SPOT-4 (20 m), and Landsat-5 (30 m)]. All three images were acquired between June-July 2006; also when the field spectral measurements were collected.
I am also interested in knowing the effect of spatial scale of the satellite imagery in its efficiency in predicting the ground vegetation (narrowband VI or LAI) for which I am using spatial regression methods at the original spatial scales (10, 20, 0r 30 m) and also at common coarser scales of 30, 40, 50 and 60 m (to compare across scales and sensor platforms).
My field site design is in a cross-shape with transects of 100 m in the N-S and E-W directions and intersecting at the site centre. Thus I have about 20 sample plots per site (5 plots in each arm of the cross) where from each plot I collected canopy spectral measurements with the ASD spectroradiometer.
My question is when I am upscaling the satellite images to common coarser scales (of 30, 40, 50 and 60 m) to compare with my field measurements, is it advisable to use the 'nearest neighbor method' or the 'pixel aggregate' function for the 'Resize Data' Tool in ENVI? I extracted reflectance for each band of a satellite image (at each spatial scale) for the corresponding sample plot location to calculate the broadband VI, and I am using this to compare to the narrowband VI or LAI. Note: I do not upscale my field VI.
Please let me know your feedback on this. Thank you very much.
--
Sincerely yours,
Arun Govind
|