h5py 3.0 changed the way in which strings are read from HDF5.
It now returns them as unicode strings instead of bytes strings.
Although I am happy that this has finally been changed, code everywhere assumes that the strings are read in as bytes, and therefore convert them to unicode.

The PRISM code can be easily changed to reflect this, but it would also force users to always use h5py 3.x, which I think is a bit too early to do.
Therefore, for now, PRISM will not support the use of h5py 3.x until the latter has released a few versions.