Our hard working friend @0709 has created a very interesting and insightful test track to expose some current shortcomings with Locus slope filtering. What it shows is a smearing of track slope resulting a deceptively lower % at places where the slope is changing. Because SRTM elevation data is only every 30 x 40 metres it may not be possible to improve the situation much but I thought I'd ask...
Does this really matter? Well say there is a short sharp upcoming hill, say +20% slope, but only for 300 metres. Locus may only show this tortuous climb as +10%.
As shown here in Garmin Basecamp (also similar in Gps Visualizer) the attached test track has precise transitions
from 0%(https://s2.postimg.org/ujgj5kwv9/20171115_-25_slope_x_20_flat.png) (https://postimg.org/image/ujgj5kwv9/) to 25%(https://s2.postimg.org/exz7lp5id/20171115_-25_slope_x_20_25.png) (https://postimg.org/image/exz7lp5id/)
Basecamp displays the slope accurately as 0%, +25% and -25%.
Locus on the other hand has quite aggressive filtering. The first +25% 'hill' starts at exactly 1.9km but the Locus chart shows the slope slowing increasing at 1.5km and only reaches the proper +25% at 2.3km.
(https://s2.postimg.org/6fprhewfp/chart_1st_25.png) (https://postimg.org/image/6fprhewfp/)
If the track style is set to manual slope coloring -25% to +25% you would expect the color to be green from 0km to 1.9km then dark red/ brown after 1.9km. But instead the color starts changing thru various shades over a distance of 500 metres, starting before the hill even begins:
(https://s2.postimg.org/62ydb9ogl/track_1st_25.png) (https://postimg.org/image/62ydb9ogl/)
One real-world shortcoming is when a steep slope is less than 500m long. For example below the 25% slope is 200m long, then flat again. Locus calculates the slope as only 15%. And the track color is pale green/ orange for only 100m.
(https://s2.postimg.org/8x1iosb85/chart_2nd_last_25.png) (https://postimg.org/image/8x1iosb85/)
Good day Andrew,
thank you for a precise description of a problem.
Unfortunately only what I may do now, is to create an new task in our "task system", that will collect problems related to filtering of measured variables with a small promise, that I'll look at it more completely during next year. Thanks for understanding.
Quote from: menion on November 24, 2017, 08:29:21
Good day Andrew,
thank you for a precise description of a problem.
@menion - many more thanks to @0709 for the detailed analysis than to me (just reporting his work).