in this case just listing and then subtracting vertical components of the start and ends or climbs, which for this simple example are also all but one at changes in direction. no need to set a sampling frequency as it’s visually obvious where the three or four climbing segments start and end.

i’ve done it before by projecting the line onto the surface created by the contour data and the using a script to sum any positive differences between subsequent points on the line. the projection algorithm introduces vertices wherever there needs to be a kink last a certain degree of variance.

it’s worth noting that the contour data is itself created by interpolation from a (very, very large) number of points, so it’s not perfect either. but it’s within inches in most of the dozens of locations that i’ve spot checked it against surveys of actual pieces of property. data like this is very hard to come by at the scale needed for something like cycling.