Ask A Cartographer

Rsampling discrete raster datasets to create smooth landcover for use in bumpmapping

November 01 2010 | 3 comments
Categories: Map Data

Hello ESRI Cartographers - I am developing a methodology for creating land-use categories for use in the bump mapping workflow. Essentially it is for maps or areas where suitable land-use data is not available. By using freely available ETM+ data (GLCF) and analysing it using NDVI (Normalised Difference Vegetation Index), I can reclassify the resulting raster into land-cover categories (5 or 6 categories - usually water, rock, soil, grass, open woodland/bushes and dense forest). I then seperate these out into vegetation masks using reclassification.

The trouble is the data is 30-60m resolution, and when I get the resulting rasters they are quite blocky, at the scale i am intending. Since the land-cover/bump mapping is to give an impression of land-cover for the reader (rather than for scientific purposes) I would like to resample down the discrete raster (may be to about 5m) and smooth it in to more pleasing, rounded LC parcels.

For the life of me I can't work out which method to use. I have tried resampling and using nieghbourhood statistics, Majority with circle 2-3 or rectangle 3, but with unsatisfactory results (areas of NoData and smaller blocks disappearing).

I have attached a Word file showing a diagram of what I am trying to do. In my example there Idouble the resolution, but in reality I would like to increase it by a factor of 10.

I would like to share my workflow with your readers once I have refined this aspect of it. I have already achieve magnificent results for smaller scale maps, this question relates to my ability to create effective larger scale maps.

Cheers.

Mapping Center Answer:

You will likely find the workflow you are looking for in online help for the Generalization toolset in Spatial Analyst -- here is a link to the online help to get you started: An overview of the Generalization toolset.

Thanks for the Help posted by Alan Boatman on Nov 4 2010 8:06PM
Thanks very much for your help. Couldn't get any of the Generalisation tools to work satisfactorly. I ended up going back to the original NDVI continuous dataset and resampling, reclassifying from there. This seemed to work well. I then used the Generalisation tool "Boundary Clean" which was marvelous and really cleaned up the resulting raster of annoying speckles and made nice smooth boundaries. Lovely tool ;-)
Thanks for sharing your results posted by Aileen Buckley on Nov 5 2010 10:03AM
Thanks for sharing your workflow and your experience with the tools! This will also potentially help others!
Rsampling discrete raster datasets to create smoot posted by Kyaw Naing Win on Nov 18 2010 3:13AM
Non of the resampling methods you tried will work for your case.

To my understanding you need to split your raster pixels into smaller ones.
Now you have 30m grid/pixels that will look blocky at some large scale. You cannot create a canopy top (a dome) of 30m size on this raster because it is represented by one pixel and that can hold a single value to represent a shade. To get a look of dome of 30m size you need enough pixels within that 30m extent. Therefore you need to populate more pixels for your map.
I hope you are familiar with raster processing environment setting in Spatial analyst tool. Set the smaller size for the "Cell size". Open raster calculator and multiply your raster file with 1 so that any value will be kept intact and out put file will be upsampled (i.e increase in pixel number.
Note: Setting the cell size to half will make four pixel out from one pixel of your classified raster file. Output file size will be four time larger than your original file. Try changing smaller size until you get a good look of bump. for testing purpose you better set smaller processing extent in the Spatial analyst option.

If you would like to post a comment, please login.

Contact Us | Legal | Privacy |