Predicting race arrivals times

I need a little advice from the physics guys.

The Nature Valley Grand Prix is part of the National Racing Calendar. It'll include road races in
Winona, MN. The pro/elite women and men will race consecutively (i.e. the women will finish
before the men start). We'll have criteriums in town while each of the pro/elite races is out in
the country.

We've allowed a pretty long time cushion betweent the finish of each set of criteriums and the
expected return of each pro/elite pack. Still, we'll want to regularly update each packs ETA during
the races so that we can adjust the length of the crits if the road racers are going much faster
than we're expecting. We'll have spotters on the course who will call in times as the leaders pass
and we'd like to use these time points to predict likely finish times (with the assumption that the
average speeds will be roughly constant).

My problem is that the road race course is not flat. The women will see three one-mile climbs and
the men will see four. There is also a lot of false flat. So I'm thinking that our assumption needs
to be that average POWER is constant, not average speed. I have no idea how to get average power
from speed or to convert this to an ETA. I realize that we'll never be completely accurate, but I
think that we can do better than we would if we assumed average speed being constant.

Any help?

David