There is not method that you can consider it to be the best, it depends on the characteristics of the plant and the conditions used for the marker. If the particular method has been used succesfully to detect variations in natural populations of your plant of interest (able to distinguish between individuals), then you can use it to detect somaclonal variation.
Although SSR can be quite useful, you need to have information about the genome sequence of your plant of interest, since specific primers are required. There are plenty of markers that can be used, you need to check their characteristics, and choose the best according to your needs. It is also advisable to use more than one marker for assessing somaclonal variation:
Of course, if there are SSR primers for your plant of interest (or for a very close relative), that would be the obvious choice. If that is not the case, use of SSR is out of the question, unless you want to isolate and develop your own microsatellite sequences. Again, there is not such a thing as the "best" method, that depends largely on your conditions.
References given by Dr Gustavo Acevedo-Hernandez ·support our finding published in Ignatov A., Mallabaeva D., Dorokhov D., Skryabin K. 2004 Molecular markers for biodiversity and biosafety assessment of main GM crops and its wild relatives. In: Genomics for Biosafety in Plant Biotechnology Volume 359 NATO Science Series: Life and Behavioural Sciences Edited by: Jan-Peter Nap, Atanas Atanassov and Willem J. Stiekema August 2004, 256 pp., ISBN: 1-58603-432-4 pp. 90 – 97
RAPD was the best for somaclonal variation, difference found by inter-Transposable Elements (ITE)-PCR had best correlation with ecotypes (cultivars), when ISSR - with susbspecies/species level. Sure, to use RAPD your must solve some technical problems: high DNA quality, optimal Ta for each group of primers, etc.