Finally Finished Keystone Trail in Omaha

Yesterday I finished the last missing piece of the Keystone Trail + Bellevue loop + Big Papillion Creek segment. It took me 19 separate hikes or bike rides for a total of 146.25km. Here is the edited (just the trails) version of my GPS tracklogs (shown in green):

Keystone-2014Jul3

Since I’ve never covered either trail, end-to-end, I’m using my route tool in Garmin’s Mapsource application to measure the distance at fairly high resolution (we’ll come back to accuracy issue) so I get 16.0km on the Big Papillion segment (west fork), starting at north end until it joins the Keystone). I measure the Keystone, starting at north end to south end (it loops around Offut to Haworth Park) at 40.7km (25.2 miles, close to the sources which range 24-27 miles). So the combined length of 56.7km means I’ve got 2.6X coverage in my tracklogs (not enough for high accuracy).

Now riding/walking these trails is no big deal. Probably many cyclists cover the entire length in a single ride and possibly a few people have walked the entire distance. But it’s still something for me to do to get out and actually use muscle power to cover some distance. Also creating GPS tracklogs and thus routes is irrelevant since these trails are well documented in other sources, plus you hardly need a GPS to follow a concrete path. So my “mapping” exercise is fairly pointless and so serves just as an incentive to me to actually cover these trails, plus think about the technical challenges of accurate mapping using commercial GPS.

Now accuracy in these measures, both the total length of all my 19 excursions and the length of the trails themselves is subject to various measurement errors. The city has put markers on the trail (usually at 0.1km spacing) so if I’d kept more records of these I could possibly get a more accurate length of the trails (assuming the city did accurate surveying). So eyeball tracing of my tracklogs into a route is probably subject to about 0.3km error (it’s a tedious process and my coverage is not good enough to get real accurate track of each trail). And using handheld GPS to measure tracks is subject to numerous errors. For instance:

Keystone-I80crossing

I have crossed (under) I-80 on the trail five times. These tracks are spaced 39m (measured where Garmin shows I-80 which is not correct location of I-80, plus I-80 is divided into two bridges at this point) and the trail itself is only about 2m wide, so a rather large spread. Furthermore using Google maps (assuming they’re more accurate) the trail in between the westbound and eastbound lanes of I-80 is shown where the yellow cross is (which is about 12m west of where I plot the route (the purple line), or as shown below:

Keystone-I80crossing(Google)

So an interesting question is how many more tracks do I need, just for this one location to get a more accurate notion of exactly where the trail is?

Now, assuming their data is good, I could, as a very tedious exercise try to create a “path” in Google Earth (the tools seem crude and easy to screw up) to get a very accurate survey, but what’s the point. Additionally, someday (in better weather) I could ride the trail end-to-end (or perhaps in segments) using both my bike’s cyclometer in addition to the GPS to attempt to get more accurate measurements (and take note of the markers as well).  That’s quite a lot to do just to get somewhat better results, but perhaps I will since it is yet another excuse to get some real exercise.

Note to self: Since making paths in Google Earth is hard, plus I need Garmin .gdx file routes, here’s a future coding exercise. Trace, at high resolution, small segments in Google Earth (so I can discard mistakes), figure out Earth’s format and make a reader that can merge the segments into a single path and then output in Garmin format so I can overlay that on all my tracklogs I’m accumulating.

Here’s a further example of the hazards of using a consumer grade GPS to do mapping:

corner

This is the map of 24 tracklogs crossing the intersection nearest to my house. The streets run exactly north-south and east-west. In my logs I stay on the street, both sides, so the tracks highlighted in pink are fairly clearly anomalous. Measured on Google Earth (challenging due to many trees obscuring the streets) the north-south street appears to be about 6.5m wide, but my tracks span about 15m (tossing out a few outliers). This is a fair amount of data (I’ll get more since I record all my rides and walks and will pass through this intersection many times).

I have a hard time thinking how I’d program a solution, not just the find the center of all the tracks (done that, fairly easy, just do linear regression constraining slope to be infinite or zero), but figuring out the actual width of the street would be tricky. So considering how to use commercial GPS for hyper-accurate mapping is a real challenge. Of course, this is not needed since this is already accurately mapped, but I can use this exercise to consider how to do mapping with GPS in order to apply that to some trail that is not well mapped (but who would care, really, and why? Just a project to exercise feet and mind).

And here (below) is some of the challenge to use Google Earth:

keystone(north-GoogleEarth)

The thin white line is where Google has mapped the Keystone, but in the northernmost part you can see the actual trail (around the open playground area) and Google’s trace is not precise. But the challenge is where the trail goes under all those trees! Google shows the thin white line but we have no visual reference to know if this is exactly right! Here’s my tracklog (just one out and back walk) and the route I plotted:

Keystone-north_end

Who is right? And how many trips do I need to make to get a more accurate trace? And when would my trace, overall, be more accurate than Google. And who cares? Except as the intellectual exercise of how to map a trail accurately, again given an easy trail like this, if we can master it, then apply those techniques to a harder trail out in the boondocks somewhere. (Or even the Camino de Santiago where the one GPS track I found, when superimposed on Google Earth, is also full of errors).

So the sure only thing this project has done is gotten me outside to cover 90.7 miles (again, subject to some accuracy limitations) via muscle power and that’s better than just staying indoors. Once the weather is more suitable for biking maybe I can get a lot more data and eventually know the precise path more accurately than any publicly available survey – oh boy, won’t the world be happy to have this bit of knowledge.

 

 

Advertisements

About dmill96

old fat (but now getting trim and fit) guy, who used to create software in Silicon Valley (almost before it was called that), who used to go backpacking and bicycling and cross-country skiing and now geodashes, drives AWD in Wyoming, takes pictures, and writes long blog posts and does xizquvjyk.
This entry was posted in comment and tagged , , , . Bookmark the permalink.

3 Responses to Finally Finished Keystone Trail in Omaha

  1. dmill96 says:

    I just completed a tedious exercise which was using the path tool in Google Earth to map the Big Papio Trail by eyeball. Since it is easy to make mistakes and editing is challenging, I did my digitizing in segments, saving after a few minutes work. It’s critical to keep the segments in the correct order as then I used an XML editor to “merge” the segments into a single segment. My tracing is fairly accurate but shadows, trees, roads (over the trail and obscure it), and some weird effects of elevation display in Google Earth certainly created some amount of error (hard to say how much). When done my completed path was 16.0km (vs 16.7 that I generated overlaying a route in Mapsource over my recorded tracklogs). The 16.0km matches some published sources more accurately.

    Now my next trick will be to figure out how to convert all this data to a .gpx tracklog so I can load this into Mapsource and superimpose on my recorded tracklogs.

    I completed 15.1 miles today so most of the BigPapio I now have quadruple coverage (out and back, twice) but a few segments are only double coverage. Once I get the Google Earth eyeball digitized data superimposed (assuming it is right) I can find places where my tracklogs are either spares in data and/or disagree with the Google Earth and get more data for those areas. On today’s ride I sometimes walked the bike in order to get higher resolution in the tracklogs but getting very high precision is going to be tricky.

    Once I have sufficient tracklogs then I face the interesting challenge of developing algorithm to do “best fit” of tracklog data to the “average”. Given the serpentine path, plus the logs go opposite directions in some stretches my head is buzzing with how an “averaging” algorithm is going to work, but this will be a good challenge.

    Of course, again, this whole project is sorta stupid since I’m getting answers to a question no one is asking and/or there are already good answers and/or no one cares, but for me I get two benefits: 1) the exercise completed to get tracklog data, and, 2) to better understand how to accomplish this task because someday I might use these techniques on trails where the answer is not already known and thus I could publish some “original” data.

    • dmill96 says:

      Oh, I misremembered. I was just checking my post and actually my route I generated in Mapsource, overlaying my tracklogs, was actually 16.0km also (thus matches the Google Earth tracing). It was my route for Keystone that had the .7 (instead of .0) in its length.

      So it will be interesting when I manage to get the Google Earth tracing superimposed on my tracklogs for the Big Papio.

  2. dmill96 says:

    And I finished another tedious exercise and now have the entire Keystone digitized from Google Earth. Its total length is a bit different, 40.9km, compared to the route I synthesized from tracklogs in Mapsource. There were parts of this trail that were hard to do due to trees or obscuring overpasses so I would believe there is a bit of error in this track, but it’s probably the best I can do. Now let’s convert it to .GPX and compare to my GPSr tracklogs.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s