expdp/impdp 1TB data from one schema user to another without creating file

If you are using Unix you can used named pipes.

If not, you might be able to use datapump import with network_link and remap_schemas and a loopback database link.

I've never used network_link with datapump imports but in theory it should be possible.

Named pipes would likely be the simplest.

From one window:
mknod -p mypipe
exp userA/password owner=userA file=mypipe


Then from another window:
imp userB/password file=mypipe full=y

How is the schema currently being backed up? Can you pull from a backup copy?

Must the source schema be active (users) during the pull?

Say more about "development is changing the owner" -- not clear.  Are you needing to replicate a production instance into a new development instance?  

Is the target schema on the same database instance (SID); on a different host; different timezone; different character set?

Subject to the above answers, my recommendation is Transportable Tablespaces. In effect, taking the logical set of data files and copying them.  They have been around since 8i, and are very easy on the system resources.  Downside, you would need to replicate the non-tablespace objects (jobs, types, etc) as a separate step.  IOW, read, and plan well.

http://docs.oracle.com/cd/B19306_01/server.102/b14231/tspaces.htm

I'll check these options out and see if they work

we are trying this one

thanks