There are different types of impervious surface in almost many of urban landscapes. Asphalt, tile roof, concrete roof, to name a few, each of which exhibit different spectrum. As with impervious surface, there may be different types of soil in a landscape just like the landscape of my interest. I am wondering would that be acceptable (in an image analysis procesure for derving land cover) getting an average of the spectra of different types of impervious to create only a single endmember for representing impervious surface.
PS: I am using Spectral Mixture Analysis for deriving the land cover of my ladscape.
Dear Mahyar,
Averaging end-members for different types for impervious surfaces is not recommended. It depends on the type of remote sensing image it is intended to use to create land cover. in addition, it depends on the spatial and spectral resolutions of the image.
The best one for extracting end-members is a high spatial and spectral resolution hyperspectral image . The end-members will be more detailed and can be refined to suit any multispectral image to use later for land cover classification.
Another best and correct method is to use hand held Hi-Res spectroradiometer
Dear M. Awad,
I am grateful for your response. Let me try to provide you with more information of my study. I am using Landsat imageries. Due to the absence of reference endmembers of my site, I would be left no option other than extracting the endmembers from the imagery itself. However, the variety of impervious surfaces and even soil types makes me wonder whether averaging them would be acceptable. Would you please advise on this condition. Thanks in advance!
Dear Mahyar
Landsat largest spatial resolution is 15 meter is this enough to differentiate between different objects such as impervious surfaces?
Another question let us assume that your pan-sharpening technique is a successful one and you can differentiate between objects especially impervious surfaces are you able to create different graphical plots depicting the spectral signatures ?
I guess Landsat 8 has more bands in the Visible wavelengths than Landsat 7 which is more useful in your case.
But in my opinion this will not work. However, some researchers use Linear Spectral Mixture Analysis (LSMA) to extract end-members of impervious surfaces check the following:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3304143/
Dear M. Awad,
Thanks for your input.
Indeed, I am also utilizing Spectral Mixture Analysis to derive my land cover. Even when I run the Hourglass Wizard, which is an automated procedure of SMA to extract the endmembers and further modelling, it returns with differential spectral signatures for different materials of impervious, most notably tile roof, concrete roof, and asphalt. Similar case is applicable to soil as most of the soil in Singapore is imported from different countries. Manual screening of the spectral plot across the landscape concurs with the Hourglass's outcome.
Suppose I have two endmembers representing vegetation and water, 3 endmembers for impervious and 3 for soils. This makes the analysis and interpretation of the results very complicated and even to some extent crude and meaningless, especially when dealing with landsats older than landsat 8, this 8-endmember model is already over the total number of bands!
And then I came across the paper whose link I am attaching in the below, which gets the average of spectra of tile roof and asphalt to produce a single endnmember for representing impervious. I thought to take same approach. However, I could not find any other example doing this. I would appreciate to have your opinion on this paper and its approach.
http://www.tandfonline.com/doi/abs/10.1080/10106040108542210
- Badges
- Science topic
- Similar topics
- Biological Science
- Bioecology
- Systems Ecology
- Landscape Ecology