You can use reclassify tool. I believe you can create up to 32 shades of color. If you repeat the processes for each raster you will have consistent symbology. If you want to use stretched symbology you need to standardize all of your raster values (e.g., -1 to 1) .

To have the consistent relevance through temporal data, you need to have max NDVI (forest, vegetation in arctic area representing the max NDVI) and min NDVI (no vegetation) showing up on your each raster. For example, if you reclassify your raster values in an area where there is only grassy area and forested area, your classification would be grass = 1 and forest = 2. Next year all forest cleared and you have only bare ground and grassy area. This time bare ground = 1 and grassy area = 2. So, your classes will be messed up. This rule also applies to standardization of the raster values.

I hope it helps.

For this kind of consistent symbology or color range visualization of raster data, I will highly recommend to use raster levelplot (http://geog.uoregon.edu/bartlein/courses/geog490/week04-raster.html) in R.

It needs RasterVis package (https://cran.r-project.org/web/packages/rasterVis/rasterVis.pdf).

May be It resolve your problem.

Hope you enjoy it.

Dear Mathieu

I would adjusted the legend with the maximum and minimum value of NDVI in the time series. Then, I would apply the same legend in all the images of the time series.

hope it works.

Regards

For image time series, use rts R package. For consistent symbology, use levelplot package.

For both analysis, please, use both packages simultaneously.

Many thanks for your answers, they are very helpful and they gave me solutions :-)

Oguz Sariyildiz , I have tried to standardize but unfortunately the results is that values do not correspond to standard NDVI and I loose the value distribution and standard deviation :-(

I had tried the second proposition, it works but it looks different as it is not stretched but I think it is ok with the data I have.

Roberto Filgueiras , in that case it works, same as previously mentioned, it does look different but I think it is the most relevant.

Oluwafemi Adeyeri Somnath Bar , thank you for your answers I need to check how to perform it with R as I have never used levelplot or rts packages.

Mathieu L. feel free to ask anything about levelplot. first install 'rasterVis' package then go to manuals of rasterVis. Below I have pasted a demo code of levelplot.

install.packges('rasterVis','sp','raster','rgdal' )

library(sp)

library(raster)

library(rgdal)

library(rasterVis)

NDVI_stack

One easy trick is merging two fabricated pixels with max and min values to your images. You can merge them outside the map extent, so that they won't present in the final map. This will keep your colour ramp consistent over the NDVI values in a series of maps.