You should have different treatments. Degraded would be control treatment and improved is another one. This is the minimum as you can have several treatments for improved sites. The soil, slope, and other environmental factors should be the same otherwise you have a problem explaining your finding (bias which you can deal with it). Then you can take measurement on productivity (biomass), species composition, percent cover, etc.

@Mounir Louhaichi thanks for your wonderful answer that fill my gap and guides me for further reading.

Just to add I would like to suggest you to add the following parameters for the proposed comparative study

Plant succession, frequency, basal area, soil physical and chemical status, regeneration status,  etc.

Takele, 

If your question is in regards to experimental treatments/controlled studies then I would defer to the answers already provided. I would add the note that degradation is a condition (i.e., a state) not a feature (i.e., a variable). Some of the variables mentioned above would no doubt be worthy indicators to include in a study, but may give an overly narrow picture of what you are trying to classify if the variables included are limited to only a few. To include all the necessary variables could be cost intensive.

Therefore I would recommend the following approach that addresses rangeland condition (e.g., degree of degradation) in a holistic manner (that includes soil, hydrologic, and biotic features), known as the 17 Indicators of Rangeland Health* (file attached). The 17 indicator variables are collectively used to gauge the degree of degradation (or improvement) compared to a baseline or historical ecosystem condition (which may need to be developed if an appropriate reference does not exist yet for your study area(s)). If there are known range sites within the same ecosystem-type and whose conditions are in good health: these would be good candidates for reference measurements prior to gauging degradation on the sites of interest. (See reference material provided for full methodology) 

*This method was developed for semi-arid and arid environments and has been applied extensively in North America, as well as China, Mongolia, and some countries in eastern Africa (see the link to nmsu.edu provided below). 

For a great text that includes many aspects of this holistic approach (as well as many examples of the variables mentioned in the previous responses), I would suggest the following book (with the link to Amazon bookstore provided below): 

• Repairing Damaged Wildlands: A Process-Orientated, Landscape-Scale Approach by S. Whisenant (ISBN-13: 978-0521665407 ; ISBN-10: 052166540X) 

Hopefully this is helpful in some way, 

-Ben

http://www.amazon.com/Repairing-Damaged-Wildlands-Process-Orientated-Landscape-Scale/dp/052166540X

http://jornada.nmsu.edu/monit-assess/manuals/assessment

To Ben's answer, I would also suggest you check out the two monitoring manuals available through the link to Jornada Research Station that Ben provided and consider checking out BLM's Technical Note 440 (attached.) 

Dear Takele, 

Great to see you working on the subject ! plenty to do ! - yes, you could still do what we did for nearly a century .... vegetation and soil condition surveys, comparison over years ...tonnes of it - done and done again .... never finished - takes years to do and the results are quite disappointing and questionable ! Use to do this ... not really satisfactory or efficient in terms of range monitoring !

Nowadays you could use the available RS and GIS tools ....plenty development these days , plenty sat images over years and all over countries, new GIS tools offer plenty opportunities and new development in monitoring rangeland condition and  rangeland health... the trap is not to stick to your screen and simply do nice images coloring ...you still do need to go to the field and do proper rangeland surveys and link your field results to RS and GIS ...this is the most difficult part !

Also, when using vegetation index, if you working in arid zones with a lot of bare soils, forget about NDVI .... useless in arid zones with sparse vegetation ..... use TSAVI and / or PVI  - for that you do need hard field data.... not only sitting behind your computer ... a most common failure and mistake from new (and old) fellow range ecologists (so easy to press a button on the keyboard and get some colors on your screen - and it has little relation to the real rangeland world in the field ).

Another useful indicator is  the Rain-Use Efficiency index .... try it !

and if you do have time see (available on my (ResearchGate page):

- A spatial desertification indicator for Mediterranean arid rangelands: A case study in Algeria, Article in The Rangeland Journal 35(1):47-62 · January 2013

DOI: 10.1071/RJ12021, https://www.researchgate.net/publication/303057047_A_spatial_desertification_indicator_for_Mediterranean_arid_rangelands_A_case_study_in_Algeria

- GIS-modelling of land-use trends: Impact of drought in the Naghamish Basin (North Western Egypt), Article in The Rangeland Journal 38(6):605-618 · December 2016, DOI: 10.1071/RJ16062, https://www.researchgate.net/publication/311681079_GIS-modelling_of_land-use_trends_Impact_of_drought_in_the_Naghamish_Basin_North_Western_Egypt

and an old one that may be of interest to you : 

- From Inventory To Monitoring In Semi-Arid And Arid Rangelands

Chapter · January 2010, https://www.researchgate.net/publication/303996439_From_Inventory_To_Monitoring_In_Semi-Arid_And_Arid_Rangelands

Do not hesitate to contact me if you need more info

Kind regards to you and Ethiopian friends

Gus 

Article A spatial desertification indicator for Mediterranean arid r...

Article GIS-modelling of land-use trends: Impact of drought in the N...

Chapter From Inventory To Monitoring In Semi-Arid And Arid Rangelands