When you attempt to build the upstream identifier tables, RivEX runs out of memory and fails with an error message suggesting you reduce the purge dictionary level. This can happen and depends entirely upon the topological nature of your dataset. It is most likely to occur when you have a large multi-threaded network draining to a single mouth.
This error would be specific to the river network so you need to reduce the purge level from 50,000 to something lower. What you reduce it to is a mater of trial and error, I suggest halving the value each time until it runs. It's trial and error as each network configuration will be unique.
o Locate the RivEX_Settings.xml file in the RivEX_Workspace folder for the network.
o Open this in notepad or your favourite text editor.
o Locate the PurgeDictionary node in the XML file, it is the third node down from the top of the file as shown below:
o Adjust the value, in the example above it had to be reset to 2000.
o Save the file overwriting the existing file.
The implication of the change is that RivEX will write out the data it is storing in a dictionary once the new purge limit has been hit. This means that the tool will succeed and not run out of memory but takes longer to execute as it is doing multiple writes to the file geodatabase.