Outlier in track log from Drive Smart 61

I am using Basecamp 4.7.1 on Windows to view a track log recorded on a Drive Smart 61. The log contains 649 points from a 2 hour 38 minutes drive of 183 km. Point 401 gives a speed of 135 KPH which I know is false - it is way too high.

I have no reason to believe my speed at point 401 was any higher than my speed elsewhere. The second highest speed recorded was 83 KPH. How/why does this one value show such a false reading? It is a concern that the data recorded can be so badly wrong; how can I rely on any of the speed data if it can be this far out?

Point 400 shows 30 KPH and point 402 shows 76 KPH. I suspect the 30 KPH at point 400 is also wrong: it is probably too low. If I take the leg distance and leg time for 400 and 401 combined it would give a speed of 76 KPH which looks accurate. So it seems to me as if the device recorded too low a distance for 400 and too high a distance for 401, with the 2 errors totalling to the correct answer but each one individually being wrong.

A bonus question on the same subject: when looking at the track log in Basecamp, how can I select all the points in the lest except point 401?

  • No, it is likely that only one is wrong in these instances and it will be the first one.  In your case it is point 400.  Larger then expected location  errors can occur when satellites come into and out of range.  Try duplicating the track log and deleting point 400. 

    The purpose of selecting points in that area is to create a summary for a section of the log.  The easiest and fastest way to do this is to require the selection to be contiguous.

    The best way to remove points to see the results is create a duplicate of the log and remove points by selecting and deleting to see the result and bringing them back with an undo (ctrl-z).  Do not do this on the original as there is a bug in some version of BaseCamp that will cause the loss of data in some of the columns.  If this happens, you will need to make a new duplicate.

    As far as reliability goes, points that are outliers are usually obvious and can be safely removed from the log.  It is unlikely to get a string of points that are contain large errors - an exception might be an area with a lot of signal obstruction.  At the same time, one can't use a single point as say proof of speeding. 

  • Thanks for the response. I have experimented with this and I see that BaseCamp does indeed re-calculate when I delete a point. Strangely, to get the calculation correct I have to delete point 401 rather than point 400 as you suggest, but now that I know this it is no big deal.

    I find your comment that "points that are outliers are usually obvious" interesting, particularly the word usually. It seems to leave open the possibility that the log can contain errors that are not obvious, and that is a concern, There is nothing I can do about it though so I guess I will try to stick my head in the sand and forget about it.

  • One that isn't obvious shouldn't create a large error though, so forgetting about it isn't a bad approach Slight smile

    Personally I would have deleted point 401 anyway if I knew that 135kph was definitely too high. 

  • I hadn't looked at it like that before but you are quite right. It is highly unlikely that a non-obvious error would ever be a significant error.

  • Well, it's never save to say always.  GPS doesn't give exact positions, a location without any error would be a rare case as it requires all the errors that are present to cancel out.  Errors outside the accuracy/precision of GPS show up as points displaced on the map and unlikely max/min speeds. At your speeds, you could also look for spikes in the speed graph.

    Both speed and time interval play a part.   Under good conditions, ~90% of points will be withing 3 meters of the the true location.  That means that adjacent points will have a distance error of no more than 6 meters.  The farther apart adjacent points the smaller the %error in distance.  GPS errors are not random, so the time interval between points will play a role on the true maximum distance error between adjacent points.