Very nice. Years ago I used to do something similar but with power and speed data: I'd calculate the virtual slope based on speed and power and then use that to get the "weighted slope" conditional on the slope > x where x is a small positive number like 1%. This is virtual slope, so it includes wind -- a headwind increases virtual slope while a tailwind diminishes it. Conditioning on x>0 removes flats and downhills, mostly: the mostly is cuz of headwinds. I thought this might be better indication of how "tough" a ride was; the same route on a calm vs. windy day would have a different virtual elevation gain. I thought a little bit about the proportion of Joules allocated to virtual climbing to total Joules used during the ride but I sorta gave up after that cuz no one understood back then what I meant by virtual elevation.