The new migration pressure overlay has proven to be extremely useful and it has allowed me to detect several issues with the migration algorithm. So I spent this weekend trying to tackle those. I ended up rewriting the function that calculates the relative value of a cell for migration purposes, and also rewrote the function that calculates how much population (or rather, prominence) to transfer from one cell to another during a migration event.

The resulting effects have been pretty good so far. Initial migrations happen at a much faster rate than before, and the migratory behavior looks much more sensical overall. Deserts and ice sheets no longer end up fully covered with tiny human populations on each cell. Though human groups will still venture into inhospitable places every now and then, allowing them to potentially cross over and literally find greener pastures on the other side. But populations that do not manage to fully cross a desert or ice sheet will eventually collapse or retreat to it's fringes, which is much more realistic than just staying there forever:

The problem that I have now is that polities are expanding much more aggressively now, regardless of their actual aggression preference. More than that, they spend a lot of time and effort migrating inside of their own territory, which might be realistic, but it's not worth the simulation cycles spent on them. So, next weekend, I need to figure out a way to scale back tribal expansion (and make aggression be more of a factor) without affecting how unorganized bands migrate. Also I have to solve a crash that occurs when the simulation tries to add the same cell to two different territories at the same time. Hopefully that is not going to be a difficult one to deal with...