Do you have training/reference data?

Please specify your set-up to get help.

Yes i do! But i do not know how to go about it. I wish to be given some step by step procedure on how to go about it

I just read your profile and I will be grateful if you could assist me since you have used MODIS in one of your publications. I am working in china at the source region

If you have training data, the thing to do is a supervised classification. To get to nwo to the principles of supervised classification in ENVI, read the tutorials: http://www.exelisvis.com/portals/0/tutorials/envi/Classification_Methods.pdf

What to choose as classification input, hence how to process your MODIS time series, depends on your aims, your study region, and other factors. Some hints in which direction you could go:

- Derive phenological/annual statistics

- Use smoothed EVI/NDVI time series

- Use smoothed reflectance time series

Anyway, I can't tell you detailed steps, because much depends on your data and aim.

Thank you Cornelius. I will try the article and procedures you have indicated.

Using only NDVI, you can get good Phenoregion maps . This will work as long as your land cover classes are phenologically distinct.

You did highlight the problem by yourself. Discriminating forests and urban areas is easy, but discriminating different forest cover types or different crop types is not that easy using a simple NDVI time series.

Anyway, I also suggest not only usins a NDVI time series, which includes the near infrared and red reflectance, but also a swir reflectance based index. The swir gives you good information on the off-seasons (e.g. soil signal), and can significantely improve the classification.