How to create a grid from 1D array using R? - arrays

Question

I have a file which contains a 209091 element 1D binary array representing the global land area
which can be downloaded from here:
ftp://sidads.colorado.edu/DATASETS/nsidc0451_AMSRE_Land_Parms_v01/AMSRE_flags_2002/
I want to create a full from the 1D data arrays using provided ancillary row and column files .globland_r and globland_c which can be downloaded from here:
ftp://sidads.colorado.edu/DATASETS/nsidc0451_AMSRE_Land_Parms_v01/AMSRE_ancil/
There is a code written in Matlab for this purpose and I want to translate this Matlab code to R but I do not know Matlab
function [gridout, EASE_r, EASE_s] = mkgrid_global(x)
%MKGRID_GLOBAL(x) Creates a matrix for mapping
% gridout = mkgrid_global(x) uses the 2090887 element array (x) and returns
%Load ancillary EASE grid row and column data, where <MyDir> is the path to
%wherever the globland_r and globland_c files are located on your machine.
fid = fopen('C:\MyDir\globland_r','r');
EASE_r = fread(fid, 209091, 'int16');
fclose(fid);
fid = fopen('C:\MyDir\globland_c','r');
EASE_s = fread(fid, 209091, 'int16');
fclose(fid);
gridout = NaN.*zeros(586,1383);
%Loop through the elment array
for i=1:1:209091
%Distribute each element to the appropriate location in the output
%matrix (but MATLAB is
%(1,1)
end
EDit following the solution of #mdsumner:
The files MLLATLSB and MLLONLSB (4-byte integers) contain latitude and longitude (multiply by 1e-5) for geo-locating the full global EASE grid matrix (586×1383)
MLLATLSB and MLLONLSB can be downloaded from here:
ftp://sidads.colorado.edu/DATASETS/nsidc0451_AMSRE_Land_Parms_v01/AMSRE_ancil/

Answer 1

## the sparse dims, literally the xcol * yrow indexes
dims <- c(1383, 586)
cfile <- "ftp://sidads.colorado.edu/DATASETS/nsidc0451_AMSRE_Land_Parms_v01/AMSRE_ancil/globland_c"
rfile <- "ftp://sidads.colorado.edu/DATASETS/nsidc0451_AMSRE_Land_Parms_v01/AMSRE_ancil/globland_r"
## be nice, don't abuse this
col <- readBin(cfile, "integer", n = prod(dims), size = 2, signed = FALSE)
row <- readBin(rfile, "integer", n = prod(dims), size = 2, signed = FALSE)
## example data file
fdat <- "ftp://sidads.colorado.edu/DATASETS/nsidc0451_AMSRE_Land_Parms_v01/AMSRE_flags_2002/flags_2002170A.bin"
dat <- readBin(fdat, "integer", n = prod(dims), size = 1, signed = FALSE)
## now get serious
m <- matrix(as.integer(NA), dims[2L], dims[1L])
m[cbind(row + 1L, col + 1L)] <- dat
image(t(m)[,dims[2]:1], col = rainbow(length(unique(m)), alpha = 0.5))
Maybe we can reconstruct this map projection too.
flon <- "MLLONLSB"
flat <- "MLLATLSB"
## the key is that these are integers, floats scaled by 1e5
lon <- readBin(flon, "integer", n = prod(dims), size = 4) * 1e-5
lat <- readBin(flat, "integer", n = prod(dims), size = 4) * 1e-5
## this is all we really need from now on
range(lon)
range(lat)
library(raster)
library(rgdal) ## need for coordinate transformation
ex <- extent(projectExtent(raster(extent(range(lon), range(lat)), crs = "+proj=longlat"), "+proj=cea"))
grd <- raster(ncols = dims[1L], nrows = dims[2L], xmn = xmin(ex), xmx = xmax(ex), ymn = ymin(ex), ymx = ymax(ex), crs = "+proj=cea")
There is probably an "out by half pixel" error in there, left as an exercise.
Test
plot(setValues(grd, m), col = rainbow(max(m, na.rm = TRUE), alpha = 0.5))
Hohum
library(maptools)
data(wrld_simpl)
plot(spTransform(wrld_simpl, CRS(projection(grd))), add = TRUE)
We can now save the valid cellnumbers to match our "grd" template, then read any particular dat-file and just populate the template with those values based on cellnumbers. Also, it seems someone trod nearly this path earlier but not much was gained:
How to identify lat and long for a global matrix?