In ERA-Interim, in high altitudes like the Himalayas or Andes one would expect there to be no data for the 1000 hPa pressure level, but ERA-Interim data contains values for these regions. Why is this?
On interpolating from model to pressure levels the data are extrapolated below the surface. For more details and algorithms see: http://www.cnrm-game-meteo.fr/gmapdoc/spip.php?article157
You can use surface pressure to mask out the extrapolated regions, though note that horizontal interpolation to a non-native grid will smear the boundary between interior points and those below the surface. Also note, that fields stored spectrally will automatically suffer this problem to some extent, even on the native grid. You can also use the surface geopotential to mask out the extrapolated regions.