Dear Maria,

The quick answer to your question is NO: it is not possible to "obtain comparable NDVI and GVI values" from multiple satellites (whether Landsat or anything else), for a number of physical and engineering reasons. These include differences in (1) sensor technology, spectral filters and calibration, (2) orbital parameters, and in particular illumination and/or observation angles, coupled with changing surface anisotropy, and (3) atmospheric correction schemes, which may rely on serious assumptions regarding atmospheric composition, amongst others.

At the core of this problem is the erroneous (though very common) approach of using of the same fixed formula for the vegetation index, which is applied indiscriminately to all satellite instruments, thereby ignoring all these issues. The correct solution is to (1) identify a biophysical parameter of interest for the target, for instance the Fraction of Absorbed Photosynthetically Active Radiation (FAPAR), (2) derive sensor-specific algorithms that deliver that biophysical parameter for each of the platforms you want to use, and (3) to verify the validity of your approach by benchmarking your results over areas and time periods when two or more of your satellites overlap. In other words, don't use a single algorithm on multiple satellites (this yields incompatible results); instead, use sensor-specific algorithms that are designed to deliver a very specific biophysical product, independently of the technical characteristics and specifications of the instrument(s).

For more information on this approach as well as concrete ways and means to compute such a parameter, consult the many publications available from

http://fapar.jrc.ec.europa.eu/Home.php

Good luck. Michel

Thank you for an answer!

I tried to go similar ways. And what about GVI? It seem to be calibrated for mss, tm, and oli sensors. Or its values also cannot be compared?

It's a matter of principle: there is no reason to believe that anyone of these simple indices are optimal for a particular purpose, and even if they were "optimized" for one instrument, there is no reason to believe they would automatically be optimal for another sensor.

If you'd like to investigate this issue further, and in particular learn about the process of optimization, you may want to look at this paper:

Verstraete, M. M. and B. Pinty (1996) 'Designing optimal spectral indices for remote sensing applications', IEEE Transactions on Geoscience and Remote Sensing, Vol. 34, No. 5, p. 1254-1265, doi:10.1109/36.536541.

Cheers, Michel.

Thank you for useful links!

It seems that I should develop something special for the region I study. But first I've got to learny carefully with exsisting aproaches.

Thank you for your suggestions. You pro

ved many of my thoughts and experience.

Maria

Well, it is an interesting discussion with good information.... Thanks to you both for raising and commenting this issue. 

If you are interested, there is another curious discussion. And one more document: (see attachment). I will find a link for downloading the full text on monday.

Discussion

https://www.researchgate.net/post/How_do_I_perform_an_image_difference_between_TM_and_OLI_vegetation_index_layers#view=54c75b7bd2fd64e64e8b4567

Thanks Maria .....I found the link of the full text.

I agree with Mr Verstraete, to compare with between values from multiple platforms, many factors must be taken into account such as morphological, atmospherical and instrumental factors

Good informative discussion Michel M. Verstraete and Maria V. Kozlova . But I have a query in this regard: Will this problem still be prevalent if we use the surface reflectance images from both TM and OLI, as I am with such outputs when OLI is giving higher NDVI values than the TM data over the course of two decades.