Adjusting for GPS elevation troubles

Posted by Dave on October 18, 2011 | 2 Comments

This weekend, on a whim, I decided to run the Bridges Half Marathon in Chattanooga. It turned out to be a pretty good whim, because I set a PR in the race!

As always, I like to take a look at my GPS record of the race afterwards to see if I might have done something differently to perform even better. But as I have long suspected, when I started to look at the elevation profile of the race, it seemed to completely botch up on bridges.

From a geological perspective, this bridge doesn't exist

Elevation data on the typical GPS trainers that runners use is notoriously bad. You can take a look at the raw GPS data from a run through Kansas and it will look like the Himalayas. To get around the problem, many running websites allow you to correct the elevations by cross-referencing against known databases maintained by NASA and the USGS.

The problem is, those databases don’t account for man-made objects. Run across a bridge, and from NASA’s perspective, you cruised across the river on your hovercraft, 100 feet below. But I hadn’t ever really come up with a clear illustration of the problem.

After uploading my data to the Garmin Connect website, here’s what my elevation profile looked like:

Garmin Connect Elevation Profile

I don’t remember the race being nearly that hilly! The last mile, for example, was completely downhill, and this profile includes a 60-foot climb! Similarly, miles 1, 5, and 7 include major dips that weren’t actually there; I was running across bridges at the time.

I decided to take a closer look at the elevation profile while simultaneously monitoring the satellite map of the course. Based on that observation, I was able to create a modified elevation profile that I think more closely reflects what I really ran:

Revised elevation profile accounting for bridges

As you can see, the course now looks considerably flatter. I also cut out two phantom hills in miles 2 and 4 that I’m quite sure weren’t there.

So what kind of quantitative difference does this make?

Take a look at this graph comparing my pace on each mile with the uncorrected elevation gain:

There is little correspondence between my pace on miles with a lot of elevation gain and the miles with little gain. I know I slow down for hills, but this graph doesn’t show how it affected my performance in the race.

Now when I adjust the elevation for the bridge problems, the effect of hills becomes much clearer.

My slower miles, like 7 and 12, now correspond to the miles with the biggest hills. There are some exceptions, but I can remember what happened on those miles as well (on Mile 10, for example, I took a little extra time at a water stop).

It’s probably not worth it to do this kind of analysis for every run, but for important races, it could help you make critical adjustments to refine your racing plans in the future.


2 Responses to “Adjusting for GPS elevation troubles”

  1. The House of Husar
    November 17th, 2011 @ 10:50 pm

    Very interesting. I will have to pay attention and make an effort to run some of our bridge in Iowa.

  2. andy
    February 16th, 2012 @ 11:34 am

    GPS devices calculate your position using radio signals from various satellites by comparing the precise time that the satellite is broadcasting.
    Man made objects like bridges and close by walls, foliage, hills, mountains, clouds, rain and even fog can and will disrupt the passage of the radio waves . The variations from the straight line to the satellite with clear sky are difficult to be accounted for and will result in a coarser accuracy.

    The usage of the GPS was intended for the military and the current accuracy was deemed enough to get a squad of armed men close enough to the designated target or an airplane close enough to the aircraft carrier. GPS was never intended to be trusted blindly.

    I have seen a doctoral research paper on the effect of perturbations in the computing of your position. It involved several off the shelf devices, a GPS device made in the lab, a fixed position determined from the topology dpt, a lot of data acquisition boards, several PC, one mainframe and several weeks of data correlated with pictures of the sky above – remember those clouds?

    If you need to pace using the GPS then you are fighting a loosing battle. You are better off using rpe, your wits and a map.