Category distinctions in the input raster map are preserved. This means that if distinct category values are adjacent, they will NOT be clumped together. The user can run r.reclass prior to r.clump to recategorize cells and reassign cell category values.
r.clump can also perform "fuzzy" clumping where neighboring cells that are not identical but similar to each other are clumped together. Here, the spectral distance between two cells is scaled to the range [0, 1] and compared to the threshold value. Cells are clumped together if their spectral distance is ≤ threshold. The result is very sensitive to this threshold value, a recommended start value is 0.01, then increasing or decreasing this value according to the desired output. Once a suitable threshold has been determined, noise can be reduced by merging small clumps with the minsize option.
r.clump can also use multiple raster maps of any kind (CELL, FCELL, DCELL) as input. In this case, the spectral distance between cells is used to determine the similarity of two cells. This means that input maps must be metric: the difference cell 1 - cell 2 must make sense. Categorical maps, e.g. land cover, can not be used in this case. Examples for valid inpat maps are satellite imagery, vegetation indices, elevation, climatic parameters etc.
r.clump works properly with raster map that contains only "fat" areas (more than a single cell in width). Linear elements (lines that are a single cell wide) may or may not be clumped together depending on the direction of the line - horizontal and vertical lines of cells are considered to be contiguous, but diagonal lines of cells are not considered to be contiguous and are broken up into separate clumps unless the -d flag is used.
A random color table and other support files are generated for the output raster map.
g.region raster=lakes -p # report sizes by waterbody type r.report lakes units=h # clump per raster polygon r.clump lakes out=lakes_individual # report sizes by individual waterbody r.report lakes_individual units=h
Perform fuzzy clumping on Landsat 7 2002 imagery (North Carolina sample dataset)
g.region raster=lsat7_2002_10 -p r.clump in=lsat7_2002_10,lsat7_2002_20,lsat7_2002_30,lsat7_2002_40,lsat7_2002_50,lsat7_2002_70 \ out=lsat7_2002_clump threshold=0.045 # reduce noise r.clump in=lsat7_2002_10,lsat7_2002_20,lsat7_2002_30,lsat7_2002_40,lsat7_2002_50,lsat7_2002_70 \ out=lsat7_2002_clump_min10 threshold=0.045 minsize=10
© 2003-2020 GRASS Development Team, GRASS GIS 7.8.3dev Reference Manual