I've been thinking for a while that geographic information software holds a ton of potential for use in political science research. So much of what we examine in elections or representation or redistricting or almost anything else has a lot to do with geography, but we don't often study things in those terms.
So I was pleased to see that two grad students at the University of Maryland have started using GIS to study party factionalism. Shanna Pearson-Merkowitz and John McTague have a piece in the new SPPQ (subscription req'd) that finds that states with more disperse populations tend to experience more factionalized parties. The argument is that in areas with a centralized population, a party can recruit candidates from just one "farm team" and avoid divisive primaries. However, if a party's voters are scattered over several disparate geographic areas, they'll have multiple farm teams that often do battle in primaries.
The article includes these nice pictures of the counties that are the sources of electoral strength in each of Florida's parties. Democrats can reach a supermajority by just drawing on eastern counties, but Republicans have to draw from the panhandle, as well, which tends to factionalize the GOP there.