Dear Gus,

Sure, NDVI is not appropriate index in most (although not all) studies dealt with arid, and sem-arid plant communities. It is greatly influenced by the background, and thus, non-sensitive to the low-coverage plant communities, especially xerophytic ones. For example, the reddish-coloured, or saline (bright, ) soils, etc., may lead to the biases in the NDVI. In addition, NDVI values often fall to zero, whereas a relatively dense plant population exists in the location.

All the methods of image processing for desert, and semi-desert plant cover research should include the correction for the background (e.g. soil), which, increases the contrast between plant, and non-plant features. But for every region different indices might be optimal, because of the different environmental conditions, available imagery, aims of the study, and other factors. Thus, PVI or TSAVI seems to be a reasonable suggestion, but there are much more options, including the combination of different image processing techniques, and analysis of in-situ data.

You may also check the special indices developed for such a scecific plant communities (see the attached files), or Tasseled cap transformation methods, etc. Just see some related publication attached.

Dear Gus,

Sure, NDVI is not appropriate index in most (although not all) studies dealt with arid, and sem-arid plant communities. It is greatly influenced by the background, and thus, non-sensitive to the low-coverage plant communities, especially xerophytic ones. For example, the reddish-coloured, or saline (bright, ) soils, etc., may lead to the biases in the NDVI. In addition, NDVI values often fall to zero, whereas a relatively dense plant population exists in the location.

All the methods of image processing for desert, and semi-desert plant cover research should include the correction for the background (e.g. soil), which, increases the contrast between plant, and non-plant features. But for every region different indices might be optimal, because of the different environmental conditions, available imagery, aims of the study, and other factors. Thus, PVI or TSAVI seems to be a reasonable suggestion, but there are much more options, including the combination of different image processing techniques, and analysis of in-situ data.

You may also check the special indices developed for such a scecific plant communities (see the attached files), or Tasseled cap transformation methods, etc. Just see some related publication attached.

Thanks a lot Maria, you confirmed our doubts about NDVI we long ago dropped when working on North African and Central Asian steppe and arid zones.We mostly used the PVI and TSAVI to our satisfaction as we always checked our results in the field with local colleagues... most often NDVI returned "bare soil - desert'' while PVI and TSVAI propely detected vegetation where we made biomass measurements . See our papers with Slim Saidi in Turkmenistan and Algeria on ResearchGate .

Regards

Gus

Deas Gus,

Thank you very much for the reference to your papers. I'll examine them.

I think, NDVI is just believed to be a universal vegetation index, for its simplicity. Moreover, presently, there are a lot of free global NDVI products, available online. It makes the use of NDVI attractive in any regions worldwide.

In the same time, remote sensing research on desert, and semi-desert vegetation commonly demands the non-standard approaches, the skills, which exceed basic knowledge on GIS software, various plant cover reflectance spectra, including xerophyres, and halophytes. Such research often quite labour-consuming. This is because something more simple may be a matter of choice, I guess. But simple things, surely not always best things for certian aims.

P.S.: I attach two more papers, describing the methods of TCT image transformation, we also found useful for arid plant cover research, and which you maybe find interesting too.

Best Wishes,

Maria

Hi Gus,

I totally agree with Maria. Sometimes the convenience in using the NDVI is greater than its own utility (NDVI = Swiss Army Knife). PVI works that tend to attenuate the soil is of incredible utility using the trend equation, which would in this case make semi-arid areas not influence both the vegetation indice. Try also to test the OSAVI, as monitoring early agricultural areas (where the soil is still exposed) has fared well in the final results.

Regards,

Carlos

Thanks Carlos and Maria for your useful comments

Regards

Gus

Two differences:

1. NDVI is sensitive to the variation of soil background. PVI may also be sensitive to the variation of soil background. But it is affected in a different way as NDVI. Soil-adjusted VI (including TSAVI) is less sensitive to soil background.

2. NDVI is over sensitive to vegetation at low vegetation density levels, in comparison with its sensitivity at high vegetation levels (saturation effects and nonlinear effects). PVI and TSAVI have more consistent sensitivity at all vegetation density levels. They have less saturation effects.

Dear Dr Zhangyan Jiang ,

Thank you for confirming details on the use of NDVI, PVI and TSAVI. I just hope that colleagues will take notes of the limitations of using NDVI on arid zones vegetation in future publications and reports and used appropriate vegetation index.

Warm regards

Gus

Gus, Salut,

Pardon pour tes collègues...

As we observed together in 2003, NDVI is simply not appropriate in most studies dealt with arid and semi-arid plant communities. It is so old Gus... .

I hope that your colleagues should carefully take note of the limitations of using this index on arid zones vegetation.

@+

Sure Slim, .... but so many are using still the NDVI on arid zone vegetation .... UN agencies, International Agric Organizations, and many others .....Useless but who cares ?

Dear all

Indeed, the NDVI continues to be used in areas with a CV

Thanks Driss for adding to our campaign of explanation on the appropriate use of NDVI and other vegetation indices such as PVI, TSAVI for arid vegetation surveys. We do agree !

Amities

Gus

Thanks Driss ...