I'm very much new to using ArcGIS and to this forum. For my thesis, I have a raster map of Indonesia where I only look at three types of land cover retrieved from the ESA-CCI dataset. These involve: croplands, forest and grassland. I merged several land cover classes representing "croplands" as one value, the same I did for forest. In other words, I basically reclassified my land cover classes.

However, some land cover classes within "croplands" involve mosaic classes, where I had to assume my own cropland fraction (%). For example, for Mosaic cropland (>50%)/natural vegetation (tree,shrub,herbaceous cover)(<50%), I assume a cropland fraction of 60% and 40% respectively.

How do I set this fraction in my map?

Since the overall objective is to quantify how much forests has been converted (100%, one entire pixel basically) into croplands and grassland in a given year.

  • With a thematic raster, I don't know if you can. You might just have to work the mosaic class into your calculations. Calculate area change in your mixed class, then multiply the value by .4 or .6 or whatever.
    – JoshC
    May 11 '18 at 12:48

At the resolution of the ESA land cover CCI map (300 m) pure pixels are extremely rare (except if you account for pixels in the middle of the ocean). This is why the LCCS-based legends makes used of mixed land cover classes.

If you want a percentage of forest loss based on the ESA CCI dataset, I recommand you not to merge the classes before computing the change trajectories, so that you can use a different ratio for each class. If you don't have higher resolution product (nor the time to visually estimate this proportion on a sample of pixels) to estimate your average tree cover per class, then the safest is to use proportions at the middle of each class. Make sure to use this proportion at a regional scale, because the average value will be very unprecise at the pixel scale.

You can also have a look at the method described in this paper or increase the spatial resolution of your forest change using the 30 m forest change dataset from Global Forest Change. For the latter, you can document the type of forest loss at 30 m based on the most likely class at 300m.

  • Thank you for your reply. One thing that I found quite strange is that the cell size of the esi cci map (all year maps) have a cell size of 0.002 x 0.002. What usually happens when you increase the resolution? What does it affect? also, I'm not sure what you meant with "use proportions at the middle of each class". I indeed separated all the mosaic classes (in forest, cropland and grassland) when I reclassified.
    – leslie.m08
    May 11 '18 at 19:00
  • Also, what do you exactly mean with ''document the type of forest loss at 30 m based on the most likely class at 300m"' ?
    – leslie.m08
    May 12 '18 at 8:50

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.