Author Topic: DEM Accuracy (Read 9440 times)

BobBailey · « **on:** November 12, 2007, 02:44:20 PM »

I did some reading on DEM to try to understand the data a little better. (Or, I tried to, I have a headache now.)

I saw accuracy numbers ranging from 4M to 50M. How accurate are the TF numbers? Do the numbers refer to the elevation or the linear spacing or both?

How granular is the DEM information? For example, I'm looking at a topo with 66ft contour lines (which I assume is converted from 20M), what is the DEM grid for this map? Am I even asking the right question?

Can somebody give the the 25-word version of DEM as applied to tracks?

tia... Bob

Krein · « **Reply #1 on:** November 12, 2007, 05:07:21 PM »

The resolution for DEM data in TF is 30 m. This is the NED dataset, downloaded from NASA's onearth server.

This is a pretty standard dataset, so I'm sure you could find some accuracy numbers as far as the actual elevation goes.

Now, things are a little different when you start applying the DEM data to tracks. Right now in TF the length of the track matters, because a DEM is downloaded to cover the extend of the entire track (this is why it zooms to fit the track). It tends to get inaccurate with tracks over 100 miles or so.

I just wrote some code to walk the track and download higher resolution DEMs in pieces. This was so I could generate good profiles for the 2,100 Appalacian Trail.

But we are planning to use a dedicated elevation server, where elevation requests are made per point, with the highest resolution.

petersfreeman · « **Reply #2 on:** February 14, 2008, 02:24:25 PM »

Hi Krein,

I'm puzzled by something. When I draw a track from Victoria to Vancouver in British Columbia, Canada and include the ferry route, the elevations of the track along the ferry route should be all zero, however they spike up to over 100 metres, when they track gets within a kilometre of an island along the route. This is a track distance of around 100 kilometres.

I also tried a track following the highway from Revelstoke to Golden over the Rogers Pass in British Columbia and the resulting profile bears no resemblance to what one would experience driving the highway. The grades come out as high as 60% with most of the grades at 20%. Rogers Pass goes as high as 8% grade. The total distance of this track is around 150 kilometres.

I've tried all different settings and combinations (Bicubic, Bilinea, Minimum gain settings from 0 to 100, Various segment subdivision etc) but it seems to have little effect.

Could you help me understand this.

Thanks,

Peter Freeman
petersfreeman@shaw.ca

Krein · « **Reply #3 on:** February 14, 2008, 04:48:42 PM »

Peter,

In general, the accuracy of the "add elevation" function goes down the longer the track, but that doesn't really account for some of the numbers you are giving.

Have you tried looking at the 3D models that are generated in the area?

As you move the center of the view (little blue sphere) around the model the elevation at that point is displayed in the status bar of the application. That might shed some light.

Otherwise, email us a track (support@topofusion.com) and I'll take a look too.

Thanks.

petersfreeman · « **Reply #4 on:** February 14, 2008, 05:47:57 PM »

Hi Krein,

I moved the 3D map so that blue ball followed my track and the elevations look right. The road steadily climbs until the last left hand 90 degree bend before Rogers Pass and then climbs to the pass and descends steeply at first and then more gradually down the final valley into Golden. The elevations reflect this reasonably accurately.

Currently I'm using the Demo version now to try out the program As soon as I get it working, I'll purchase it as I am doing the Tour du Canada cycling trip this summer and and I want to get profiles of the trip.

I'll email the file for you to look at and I'll leave the blue ball at Rogers Pass.

Thanks for your quick response.

Cheers,

Peter

petersfreeman · « **Reply #5 on:** February 27, 2008, 12:09:26 AM »

Hi Krein,

I've just purchased TopoFusion Pro. In the meantime, while TopoFusion is getting the Profile bug fixed that Scott is working on, I need to get profiles and have downloaded 30 arc second DEM data (1 kilometre), however it is too coarse for my highway through the rocky mountains (Revelstoke to Golden, British Columbia) so although it shows the highway climbing to Rogers Pass and then down to Golden, it fluctuates wildly as some of the points are up to 500 metres off the centerline of the road, which puts them way up a mountain cliff. I need to use the 90m DEM (3 arc second).

I downloaded the 90m DEM however the grid is 6,000 rows by 6,000 columns. I am processing the data in MS Excel which has a 255 column limit.

Can you tell me where I can download 90m DEM data in smaller chunks? Can you also tell me if there is an online source where I can provide coordinates (Latitude and Longitude) and it returns the elevation (perhaps adjusted to fit within the four surrounding elevation co-ordinates)?

Thanks,

Peter