search

ncss-tech / soilReports

An R package that assists with the setup and operation of a collection of soil data summary, comparison, and evaluation reports. These reports are primarily used by USDA-NRCS soil scientists in both initial and update mapping.
15 stars 5 forks source link

add NLCD impervious surface map to reports #69

Open dylanbeaudette opened 7 years ago

dylanbeaudette commented 7 years ago

NLCD 2011 Percent Developed Imperviousness (741MB) The 2011 percent developed imperviousness layer for the conterminous United States for all pixels.

... downloading now.

dylanbeaudette commented 7 years ago

The data look good. Range is 0-100%. Need to include the metadata with the RO2 files and send to Russ / Jennifer.

Using the data opens up a new issue--summary of 0-inflated raster data:

image

A couple of options:

  1. set 0 to NULL in source rasters
  2. set 0 to NULL in raster samples
  3. use a non-Gaussian smoother in the density plots
dylanbeaudette commented 7 years ago

More ideas:

Are there other data that may result in the same kind of degenerate density estimates?

brownag commented 7 years ago

Consider reclassification of samples and categorical summary on reclassified values?

NULL-ing out the zeroes would allow us to get an estimate of the distribution of the non-zero portion... but the distribution and relative abundance of zeros is meaningful when summarizing these impervious surface percentages on a delineation/map unit basis.

Reclassification to pull out exact zeros as the "absence of impervious surface" class and then several less-pervious classes (e.g. 0, 1-25, 25-50, 50-75, 75-100%?). Then, categorical raster summary would give us the expected values for each of these classes, which should approximate e.g. component percentage of "Urban land" or impervious surface etc.

An issue to consider is that the process that determines presence/absence of zeroes is not particularly smooth, especially for example in our area ofthe not-that-developed Sierra Nevada foothills and mountains, In that respect constant density sampling may not be the appropriate way to go collecting raster samples. Though, perhaps the caveat would be that most map units in their "natural state" would have a low percentage of impervious and the summary of impervious surface is really only good for comparing MUs with a decent amount of impervious.