(Website and Android) The website estimated height gain/loss seems to have become materislly inflated compared to previous results, and with differing results to the Android app.
eg
- See route name Durham - Bishopton circular saved as public just now. This used to show as a 1290ft ascent/descent, but now gives 1800ft. Google gives 1306 ascent, 1293 descent; ie v close to the previous result from CT. On the basis of the difference in height between the foot and top of each identiable ascent, as per the elevations indicated in CT, the ascent sums to 1158ft (with other minor oscillations not yet accounted for), which seems to support the previous CT estimate, and to be incompatible with the total currently given in CT website. (Planning the same route on the Android app give ascent 911ft, which does not aseem right either ... ).
- Stanhope (Co Durham) - Hexham railway station with a via point at Allenheads (excellent cafe!). Desktop app reports as total 2100ft ascent. However, planning the route in the Android app gives the total ascent as 1991ft. Noting the ascents in the website version of the route profile, in the same way as above, they sum to 1900ft. ie very compatible with the Android app result. (Puzzling, as I had assumed it was the same underlying engine...) Google gives just 1791ft ascent.
Is there perhaps something now going wrong in the totalling of the separate ascents in the latest version of your algorithm for the website??
Comments
Having scratched my head a lot over this, I think this is a result of cycle.travel now using (much) higher resolution elevation data.
It’s now picking up small climbs that it might not have previously noticed. For example, on Durham–Bishopton, it notices little ascents before Bishopton, Redmarshall, and Sherburn. I’m sure these are all real ascents – you can see them if you follow the route on Street View – but it’s always difficult to know at what point to include these in the total ascent calculations.
As ever, there isn’t really any “right” answer when it comes to elevation and comparing numbers across sites doesn’t really give much illumination. Google appears to be much lower than a couple of cycling sites I’ve tried the same routes on. But I’ve just done a couple of little tweaks to the c.t thresholds which might make the effect less stark.
(There’s also a particular issue with Durham–Bishopton in that the rail trail between points 10 and 11 should really be a solid climb: currently it’s showing undulations from the surrounding land. c.t is generally smart enough to know that rail trails have gentle, consistent climbs, but unfortunately the UK is blessed with one particular OpenStreetMap contributor who keeps deleting this information, so it doesn’t always come through to c.t as it should. I’ll reinstate it in OSM when I have a spare minute.)
I tend to develop stuff on the website first and then move it to the apps after a while, because it’s much simpler to experiment on the site – you don’t have to wait for Apple or Google’s approval every time you want to publish something! Consequently the algorithms usually lag a little. They are also (by necessity) written in the different programming language for each platform so there’ll be a few implementation differences too.
Richard,
I certainly appreciate that there is no definitive "right" result; I just wanted to draw to your attention in cases omething was going wrong.
I'm always conscious of every ascent, and my route-planning has a close eye to minimising rises; so v happy to have them noticed by CT! And few of us will complain if the records of ascent on our rides are now more flattering ...
I also see that the Bishopton route is now showing as 1600ft ascent, not 1800ft, and the Stanhope-Hexham route as 2000ft not 2100ft - I guess as part of your continuing fine-tuning you have turned the dial down a bit? Both seem more reasonable. And if/when refined OSM data that reflects the more even slope of the railtrail gets re-included, that will tend to reduce the undulations in that gradual rise, and bring the Bishopton result closer to the previous number reported by CT.
Many thanks, its a great site.