I know you explicitely asked for an ArcGIS solution. However, I just stumbled accross the same problem when trying to import GIMMS NDVI 3G data into R. So in case anybody needs this in future times, here is the solution I came up with.
Basically, my approach is built on read.ENVI from the caTools package. However, to successfully import the binary data into R, you need to set up a companion header file as mentioned above. Following the documentation provided by ?read.ENVI, it looks quite different from the one ArcGIS can deal with:
ENVI
description = { R-language data }
samples = 2160
lines = 4320
bands = 1
data type = 2
header offset = 0
interleave = bsq
byte order = 1
You need to save these lines to a separate *.hdr file, which you can supply to read.ENVI later on. Of course, if you are dealing with multiple GIMMS binary files, you can use the same header file again and again since all settings remain unchanged. Like that, it is quite simple to import the data into R:
library(caTools)
read.ENVI("/path/to/gimms_file", "/path/to/header_file.hdr")
read.ENVI returns a 'matrix' by default, which you can then easily transpose calling the t function and, finally, insert the values into a 'RasterLayer' object using raster(...).
Update:
You may now also have a look at the R gimms package which renders unnecessary the use of read.ENVI. To create an ordinary GeoTIFF from the downloaded GIMMS binary data and make the file available in R as 'Raster*' object, simply run rasterizeGimms. A comprehensive introduction to the package is available from GitBook.
