Basically, individuals out along the wave front are individuals that tend to disperse. There they are, looking for mates, and boom! the other ones out there are ones that also had that urge to disperse. And then voilà, their offspring? You guessed it, they are super dispersers.
We tested these ideas using Tribolium beetles as a model system. We put them into constructed landscapes, and let the populations disperse and grow for 8 generations. One set of populations did so without disturbance (other than censuses and getting fresh food). They could thus naturally develop the spatial population structure that makes it so that individuals that disperse most are found at the expansion edge. Another set of populations was shuffled each generation at census time, so that they could expand and spread but without developing spatial population structure. We found that yes indeed, individuals at the expanding edge of a population disperse more than those from the range core.
However, although the mean change is for population structure to lead to faster range expansion, there was more variation in the structured landscapes than in the shuffled ones. This is partly due to reductions in fitness at the expanding front. That (evidence points to gene surfing) is a post for another day. But the haphazard combination of faster but variable movement with lower and variable fitness, leads to more variable outcomes all together.
So, we can predict, reliably, that evolution of population structure increases invasion speed and makes it more variable. Understanding the magnitude of the variation is part of prediction! I said we want to predict with confidence... We may have a large prediction interval, but we can be more confident in the upper and lower bounds now.
Science. It rocks.