Are there any indices which can specifically extract barren areas only (with a precision of 60% or more) or is supervised classification the only reliable way?
First question, are you using moderate resolution optical (Landsat, IRS-P6, Spot), coarse resolution optical (MODIS, AVHRR, MERIS) or radar? Depending on which, there are a number of approaches you could take. The simplest approach would be to take an image from a wet period, calculate an NDVI image, and then threshold to extract low vegetation cover (but >0 to exclude water) and assume that anything with low veg cover during wet periods is barren or un-vegetated. There are more sophisticated multi-temporal approaches that can be applied (based on using a similar thresholding technique applied to a series of images over time).
I use Landsat 5 image. The method suggested by you is a good one and we have tried approaches similar to that, but what I want to know is, as there are lots of indices developed specifically for vegetation, water etc. are there any similar ways in which only a barren area can be extracted...
Hi Srijith, there are bare soil indices that use the short wave infra red wavelengths to identify bare soil/exposed rock, these may suit your purposes. The traditional approach is the tasselled cap approach
The trick in some ways is in terms of definition, if by 'barren' you mean an area where no vegetation grows, then low rainfall, low fertility soils, exposed rock, claypans, salt scalds and salt lakes would all qualify as 'barren'. If you're chasing one of those in particular, you'll need to focus on the spectral characteristics. Otherwise an NDVI thresholding, tasselled cap or fractional cover approach is the most defensible.