I have to classify land use land cover data for western ghats- 1,07,770 sq km area.The landsat images for the area cover 15 tiles . I have to use minimum cloud cover images . But for the same the images available at minimal cloud cover are from different dates . Can i mosaic these images of different dates (say dec1 2014 , 30 oct 2014 ) to achieve YEARLY lulc without any problem ?

  • Define 'problem'. When I did digitizing we had source imagery with several different dates tiled together. The further apart the date, the more issues with feature mismatch. A couple of months, probably not so much (unless say, all trees dropped leaves in that time or the ground became snow covered). At most it was typically something noted in metadata (ie, all features derived from images taken between x and y, or in specific cases where we had to fall back to a secondary source a particular feature might be sourced at the attribute level to that source as mentioned in metadata).
    – Chris W
    Mar 25, 2015 at 7:06
  • I think that the problem depends on what you want to do. If you want to see land cover differences between months it might be no problem(but you have to be careful with seasonal spectral signatures of vegetation). If you wanted to analyze phenomena like urban areas, perhaps you would have important differences because urban changes very fast. Land cover is kind of "stable" between months (e.g the forest dec. 2014 would be forest in apr.2014). "Problem" I think you would have if your images were from different satellites.
    – geo_dd
    Mar 25, 2015 at 8:08
  • Given that you are dealing with western ghats(of India, I presume), much of the vegetation is seasonal, and as such you will have problems if you try use data spread over a long period. It's best that the data be temporally close, but if it is not available, you need to study the weather conditions during that time, to check if there is much variation. Mar 25, 2015 at 8:48

1 Answer 1


The main issue with using scenes from different times of the year is seasonality / phenology of the vegetation.
The variation in spectral response of the vegetation across the year can make it difficult to create a cloud free mosaic, but if it is possible to create multiple mosaics over the year, that can certainly help improve your LULC as the phenology is very useful for classification.
One thing that you can try is a pixel-by-pixel NDVI maximum value composite, where you retain the regular bands in your image stack. This should be done on atmospherically corrected data - an option would be to use the PROVISIONAL LANDSAT 8 SURFACE REFLECTANCE PRODUCT from the Department of the Interior U.S. Geological Survey. It may not be perfect in your area due to known issues in mountainous regions, but it may be worth a shot.

  • Another useful information to evaluate is the phenology of the crops in the study area. Use a crop calendar to undesrtand in which stage are the main crops. If the focus is in agriculture, the best images corresponds a one or two weeks before the harvesting where the growing period is at the top Mar 25, 2015 at 11:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.