Also, if it's climate data (i.e. the results of physical measurements), you almost certainly don't need double precision in the initial data file--single precision would cut its size in half. You may want to do the subsequent calculations in double precision, of course.