Advice on reducing topographic-induced illumination on Spot images

[Shaenandhoa García-Rangel]

Dear all

My name is Shaenandhoa Garcia Rangel. I am a PhD candidate at the
Wildlife Research Group, University of Cambridge, working on a project
to model Andean bear (/Tremarctos ornatus/) habitat use across the
north-east end of the Venezuelan Andes.

I am currently working with Spot 2 and 5 images to develop a land cover
map of my study site, and and I have carried out a topographic
correction following the VECA algorithm (Gao and Zhang 2009), to reduce
the effect of the topographic-induced illumination of the images prior
to carry out a supervised classification.

I have a few questions regarding the relationship between the original
image and cos (i), as well as the correction outputs that I very much
hope you can advice me with.

1) The regression coefficients for each of the original images versus
the cos i were <0.2 in most cases and only up to 0.4 for all of them. I
wonder if this reflects that my images are not very much affected by
errors of topographic-induced illumination, despite my area being very
mountainous and if so, how can this be? and if given the low values of
this coefficient is it still necessary to run this correction before the
classification?

2) On the other hand, could this possibly be reflecting a problem with
the DEM? The DEM I used came from the STRM-90 from which I extracted the
area of my study site and then rectified it before using it for the
correction, so I am confident about the geo-reference. However, I had to
resample from 90x90m pixels to 20x20 and 10x10m given the resolution of
my satellite images, and I wonder if this might be causing the low r
coefficients for the initial correction, and if so how could I get
around this?

3) After the VECA correction was carried out I compared the result of
the regression with the ones mentioned previously. For all the images
the regression coefficient is reduced, as well as the value of the
slope, indicating that the VECA correction has in fact reduced this
topographic-induced error. However, Band 3 (NIR) shows overcorrection
given the negative values of the slope. Here, my question is, how much
is too much overcorrection?. Original images of Band 3 report the
highest topographic-induced errors (higher r and m values) and also the
highest reduction of the error (lower r and m values) but this reduction
is negative (e.g. m original image = 27.03, m corrected image = - 1.21).
The coefficient of departure for this band across all images is slightly
(the 2nd decimal point) higher than 1.

4) Spot 5 images, report very slight increases on the standard deviation
(around the 2nd decimal point), this is reflected by slight decreases
(3rd decimal point) on the coefficient of departure for one of the
images and by CV increases for all bands (except Band 3). Despite this
the regression coefficient were reduced after the correction. In this
case would it be better to use the original images for classification or
given the reduction of r I should be using the the corrected images
instead?


I would really appreciate any advice you can offer me on this questions.

Thank you very much in advance

Shena

Shaenandhoa García Rangel
Wildlife Research Group
University of Cambridge
_______________________________________________
gislist mailing list
gislist@...
http://lists.geocomm.com/mailman/listinfo/gislist

_________________________________
This list is brought to you by
The GeoCommunity
http://www.geocomm.com/