[R-sig-Geo] conservative continental "reprojection" for lon-lat input?

[Tom Roche]

http://cran.r-project.org/web/packages/gstat/vignettes/gstat.pdf
> Package gstat assumes that data are projected, i.e.
> [data] should not be provided as [latitude]/longitude.

How should a continental-scale reboxing project best cope with this
restriction? Why I ask:

I'm a student attempting to create 3D N2O initial conditions (ICs) and
boundary conditions (BCs) for a run of an Eulerian atmospheric model
over North America, specifically

https://github.com/TomRoche/cornbeltN2O/wiki/AQMEII-North-American-domain#wiki-EPA
> projection: LCC (Lambert Conformal Conic)
> standard parallels: 33°, 45° (N)
> lon/lat projection center: -97° (W), 40° (N)
> domain origin (from projection center, in km): -2556 (W), -1728 (S)
> horizontal grid spacing: 12 km
> horizontal grid count (x,y): 459, 299
> vertical grid count: 34
> max height: 50 mb

I'm attempting to create IC/BCs for my model using, as input, the output
of a more coarse-scaled 3D global inventory: 2.5° lon x 1.875° lat, with
vertical layer heights more resembling those of the output domain
(though the layer positions differ). This "global input" provides an
approximate value for the mass concentration of N2O in the air in each
of its voxels, or boxes.

So I'll need to "rebox" the global input onto the local domain: i.e.,
for each box in the North American projected domain (i.e., for the
"local output"), estimate its mass concentration based on those provided
for the global input. I was hoping to do this directly, in the manner
that raster::projectRaster allows one to "regrid" from lon-lat to LCC.
But direct reboxing from lon-lat appear problematic with package=gstat,
however. Am I missing something?

If not, I'm wondering: what would be the most safe, conservative way to
impose a projection on the global data in order to input it to gstat?
E.g., could I

1. reduce the extent of the input from global to a "subglobal" area
   somewhat larger than my target domain (which is -130 - -59.5° lon,
   23.5 - 58.5° lat)

2. impose some CRS on the subglobal data such that the areas of the CRS
   gridcells equal or closely approximate those of the subglobal grid.

? If so, which CRS to use?

TIA, Tom Roche <Tom_Roche at pobox.com>