[R-sig-Geo] Problems with extracting data from GRIB files from "The NCEP Climate Forecast System Version 2 (CFSv2)"

[k_jacobi at surf2000.de]

For a model on estimating flows from snow melt, I need forecasts on
temperature and precipitation. Forecast data (12-hourly for the next 9
months) I find on ftp://ftpprd.ncep.noaa.gov/pub/data1/nccf/com/cfs/prod/,
The data are in grib2 format, and I can load each file as a rasterbrick
and extract forecasts for certain coordinates c(x,y) like that:

tmp_brick <- brick(gribfile)
t <- extract(tmp_brick, cellFromXY(tmp_brick, c(x,y))) # time series

Although the coordinates of my raster brick are bit strange I can live
with that:

> tmp_brick
class       : RasterBrick
dimensions  : 73, 144, 10512, 559  (nrow, ncol, ncell, nlayers)
resolution  : 2.5, 2.5  (x, y)
extent      : 0, 360, -92.5, 90  (xmin, xmax, ymin, ymax)
coord. ref. : +proj=longlat +a=6371200 +b=6371200 +no_defs

Unfortunately, the above mentioned url contains only forecasts for the
last 8 days. In order to verify the suitability of the data, I need to
precess historic forecasts (reforecasts).

I found something which looked initially promising under the url
http://nomads.ncdc.noaa.gov/data/cfsr-rfl-ts9/

I can load data as a raster brick, although coordinates are even more
strange from point of view of a GIS user:

class       : RasterBrick
dimensions  : 190, 384, 72960, 1200  (nrow, ncol, ncell, nlayers)
resolution  : 0.938, 0.9473684  (x, y)
extent      : 0, 360.192, 88.32963, 268.3296  (xmin, xmax, ymin, ymax)
coord. ref. : +proj=longlat +a=6371229 +b=6371229 +no_defs

Also, I received the following warning from R

dec_jpeg2000: Unable to open JPEG2000 image within GRIB file.
Is the JPEG2000 driver available?

When I Try to extract values for a selected coordinate, the warning
repeats for each raster layer and the results either are either a vector
of 0 (precipitation rate) or strange figure around -60 for temperatures.

BTW, I can load and visualize the data with Open EV_WD (but trying to
export them has no effect).

Any help how to overcome this problem (including hints to data sources
that are easier to process) would be highly appreciated.

Klaus Jacobi