I estimate at 90F a decrease in miles/kw-hr of .02, at 115F it should be 0.05 (again my estimate with an average speed of 60mph). Higher speeds will lower this number (more energy per hour but less energy per mile) still 40% less than the impact of heating the car when it's 20F.
OK, I understand your analyses. I think your point vis-a-vis, the efficiency impact and how it varies with speed is a valid point. That said, I think we need to scope the bigger problem first. Specifically,
We want to establish a pseudo-quantitative understanding of the degradation of mileage driving in hot weather by building a reasonable model. To that end, we need to identify the heat sources and the required AC unit.
Without going into the absurd details of doing a finite-element model, perhaps we can do a simplified lumped model as follows:
> Heat Sources:
[1] conduction and radiation from the sun's rays that heat the cabin (i.e., car at rest, no cabin occupant). Perhaps we can treat the cabin as a glass greenhouse?
[2] heat from the airflow at XX-mph over the passenger compartment that brings additional (in addition to the sun's radiation) heat flux into the passenger compartment, and
[3] heat emanating from the inside of the car (electronics, etc.) and by the occupants.
> Required AC unit:
> once we have the above, we can calculate the "steady-state" heat load and the BTU required to maintain a given steady-state temperature taking into account the efficiency of the AC unit.
> and from the estimated BTU above, we can estimate the resultant degradation of the mileage efficiency.
Perhaps there are aspiring physicists who want to tackle this problem. I know I am not qualified!