All of the arrows on the food web are currently stored as individual image files for each species. This means that, to modify, add, or delete a species requires modifying the image file for every species' image that species has an interaction with. This is a prohibitive amount of work, and directly goes against the long term set of goals of having an extensible or replaceable set of species that would be appropriate for different ecosystems.
The suggested solution is to parse the text file containing interactions between species that the simulation itself uses and use this to programmatically draw the arrows between different species. The location of the different species on the food web is still specified manually in terms of pixel coordinates within a text file, so the programmatic code doesn't need to arrange the whole food web, merely draw in the appropriate arrows.