Monday, May 16, 2011

How much noise does GPS correction remove?

A few weeks ago I wrote about how elevation measures on GPS devices are notoriously inaccurate. Correcting the data based on more accurate geographic surveys can fix some of problem, but can also introduce error. For the most part, however, the corrected data seems to be better.

But when I'm running a new race, I frequently use Garmin Connect's Explore feature to see if anyone has shared their recording of the race route. It can be tricky to find an older route (hint -- focus the map on the start of the race and enter the date and approximate mileage of the previous year's run), but when you do, it can come in very handy! In less than three weeks, I'll be running the Steamboat Marathon, and I was able to find this record of one runner's race from last year (Congrats on your first marathon, bhatton59!).

Unfortunately, when you're looking at someone else's route, GPS elevation correction is disabled. That means the cumulative elevation gains reported for the route are likely to be off by quite a bit. But how much? It's hard to know exactly, as each location has its own GPS problems (trees, buildings, mountains, and so on), but I think I have come up with a couple ways to guestimate how much error there is in uncorrected elevation data.

I took my six longest runs (all over 19 miles) and viewed the corrected and uncorrected cumulative elevation data for each. For these runs, the difference between corrected and uncorrected data ranged from 171 to 626 feet. That's a pretty big range! Even when you figure out the difference based on the average correction per mile, there's still a big range. For the Fellowship of the Idiots run, the correction per mile averaged only 8.9 feet, but for a 20-mile training run I did in February, the average was over 32 feet -- that's nearly three times the amount of correction. Overall, the amount of correction ranged from 12 to 42 percent.

So how does this affect bhatton59's elevation data from last year's Steamboat Marathon? Her uncorrected record shows a cumulative loss of 2,128 feet and a gain of 742. Now since we know the race starts at an elevation of about 8,128 feet and descends to 6,728 feet, the true cumulative elevation loss must be at least 1,400 feet. So there's about 728 feet of wiggle room in the record. Based on my data, that 728 feet could be off by anywhere from 12 to 42 percent -- from 87 feet to 305 feet. That doesn't change the cumulative elevation loss much -- it's going to range from 2,041 feet to 1,823, but it means the actual uphill on the course could be anywhere from 664 to just 437 feet -- that's a big difference from the uncorrected 742 feet of gain reported on the Garmin site.

Another way to guesstimate the uphill on the course is to look at bhatton59's player:

Garmin Connect actually allows you to click on any point along the graph to get the elevation and distance information. In this way, I could click on the obvious major hills and avoid data points that look like they may be noise. Adding all these points up, I came up with a cumulative loss of 1,961 feet and a gain of 563, which is right about the middle of the range predicted by my initial calculation. That's a lot of downhill, but actually about the same as the corrected total elevation loss I recorded at Big Sur: 1,863 feet. So I should probably be able to handle the downhills, and the total of around 563 feet of climbing shouldn't add too much difficulty. My major worry will be the hill from Mile 20.23 to 22.27: a climb of about 186 feet, on tired legs. There was a similar hill in a similar spot at Big Sur, and I slowed from an under-8-minute-pace to a 9-minute pace, and never recovered. Hopefully this time around, with better hydration, I'll be better prepared for a hill in the closing miles of a race.

Details of yesterday's workout are below:

Yesterday I met up with DARTers for a 10-mile workout, my longest since the marathon. I ended up running with Tim, Terry, and Randy. In the spirit of Friday's downhill workout, Tim egged my on every time we hit a downhill stretch. I had a strong run, with an average 8:02 pace, and felt good at the end, no trace of tired legs.

No comments:

Post a Comment