Calculating the percent of binary values in grid of 1 km by 1 km across the entire raster

Question

I have a raster of the entire united states with 30 meter by 30 meter pixel size. Each pixel has a value of either 0 or 1, with 1 defining a certain type of landcover.

I need to find the percent of 1's within 1 km by 1 km areas gridded across the entire raster so that I know for each 1 sq km, what percent of it is the landcover I'm interested in.

Despite a lot of googling, I am unsure about how to proceed. Mostly how to create the grid. It looks like once I have that I could use v.rast.stats in GRASS. But maybe I'm way off.

As a result, I don't have a specific error I'm encountering, but rather do not understand the tools well enough to know how to proceed to get my desired output.

Answer 1

Since you mentioned GRASS in the question (although not in the tags), i can suggest an easy procedure. Your 30x30 raster is binary, so summing all pixels in a 1kmx1km grid will give you the number of desired landcover pixels in the coarse grid. You can do this by resampling in GRASS. If you then do

(sum/(33.3X33.3))*100

divide by the total number of original pixels in each coarse grid pixel and multiply by 100, you'll have the percent.

Here's the code in GRASS: (assume your binary raster is called lc30, and you are using a projected CRS)

# First Set computation region to 30 meter landcover raster
g.region -ap rast=lc30
# reset the resolution to the coarse 1km resolution
g.region res=1000
# Resample to the coarse grid, using "sum"
r.resamp.stats input=lc30 output=lc1km method=sum
# and calculate a new map with percentage 
r.mapcalc "lc1km_pcent = (lc1km / (33.3*33.3)) * 100.0"