Dem Grades & Accuracy

Seane2c

New member
So I was fortunate that someone who follows me took the time to create a DEM for me (Canada, Niagara Region, and I got to see first hand what this was all about. I believe it was 30M data (the NASA stuff whatever that is) in TransDem Visual Earth zoom level 15. Now I have to admit the Niagara gorge does look beautiful, and I was even able to due a mocked version of Niagara Falls to my surprise. Going into smaller bodies of water however like the Welland Canal I found the accuracy was off. I assume this is because it's 30M thus the points are registering at the wrong spot etc. So a bigger body of water (though having some faults of their own) is not as noticeable.

But having said that the grades as I've heard can be like a roller coaster (probably for the same reasons). I've seen advice such as do not try and flatten the terrain but work with it for now etc. and take baby steps. As much as I've seen tutorials on TransDem (though few) I haven't seen much when it comes to mastering the post process of DEM.

The pros of this was my entire area now had track though it needs fixing it's all there everything including ancient track long gone (I'd rather see more than less so I can appreciate that more. But the grades and understanding the best approach would be interesting. Also I've thought about getting TransDem to try and find something in Canada with even more detail (if it's possible) but the price of TransDem for 1 railroad is steep, and then add the cost of buying data from the government I guess?

Any advice or tutorials beyond this process is appreciated. For now I'm torn between doing it by hand and continuing my progress, or working out the kinks of the DEM and trying to live with this.

Thanks

Sean
 
If a gradient is one huge ridiculous, up, and down, leaps, and dips, rollercoaster, with grades much greater than 0.50%, you will usually have to fudge the grade, and apply Goggle Earth track elevation metric heights, and apply "smooth spline" to flatten and cut the terrain. Rarely does a gradient surpass 1.00%. Most of my gradients are @ 0.05% to 0.25%, in some rare places 0.50% to 0.75%, except in the mountains, where it can be as great as 1.75%.
 
If a gradient is one huge ridiculous, up, and down, leaps, and dips, rollercoaster, with grades much greater than 0.50%, you will usually have to fudge the grade, and apply Goggle Earth track elevation metric heights, and apply "smooth spline" to flatten and cut the terrain. Rarely does a gradient surpass 1.00%. Most of my gradients are @ 0.05% to 0.25%, in some rare places 0.50% to 0.75%, except in the mountains, where it can be as great as 1.75%.
Unless you model the Uintah, with 7.5% grades. I did a dem for that, but the topo maps in that part of the world aren't that great, I found it was mostly mild hills that suffered the most. There was up to a 10 meter difference in places. I cross referenced with google earth which isn't that accurate either and then referred to the book on the Route which fortunately had ruling grades included. The best thing is to have multiple reference points to rely on and you will probably get close to the original.
 
Yes, there are many exceptions, where Shays, Climax, COG, and other mountainous locos were used in the hills and mountains, and Saluda and in Europe. I was referring to the normal rolling hills mainline gradients.
 
Now I did hear some place warning not to use smooth spline on a DEM terrain because it cuts into everything else around it. Since it was just something I heard on an older forum post perhaps it's not as relevant now. I could always undo if I don't like it I suppose.

Thanks

Sean
 
Hi, Sean.
First, my experience with DEMs is not as extensive as some on here, but if you go that way, I would suggest the following:
1) Make backups of your route as you work. If you really do something wacky that you can't "undo", you can always roll back to a previous version. (I label mine "Route X-001", -002, etc)
2) If you make a cut or a mound in the landscape using the "Smoothing" tool; it can work well, but plan ahead, and use something VERY narrow - a walkway, path, etc - if you use the tracks or a road, the effect gets very wide. It also gets wider if you click more than once.
3) If you are doing a prototypical route, check Google Earth or maps to give you some guidance on how the track might be running in real life. Even on a fantasy route, this can be a good template for what to watch for.
Good luck, and have fun!
Ron
 
Now I did hear some place warning not to use smooth spline on a DEM terrain because it cuts into everything else around it. Since it was just something I heard on an older forum post perhaps it's not as relevant now. I could always undo if I don't like it I suppose.

Thanks

Sean
Yep, it will flatten the terrain in quite a wide area, it all depends on what is close to the track , if the ground within whatever grid size you using has no cutting or rise in ground, no issues, but if you are next to a creek or hill,smooth spline will cover or eat into the terrain .
 
Yeah I think I can be careful and utilize versions as well as CTRL-Z to keep things in check. I was using Basemapz before and they were fairly accurate but out by about a meter or two per board. It's also pretty useless when the train changes as it wants to find the biggest flat area to sit on making it useless for DEM. For some reason the data I was provided had more dem then the actual map (UTM included with the DEM data) showing on the outsides. Larger bodies of water are more accurate, but the Welland Canal seems way off for some reason.

Thanks

Sean
 
If you do not use the "Smooth Spline" tool (which flattens a 20m grid area on each side of the tracks), your track will be floating in spots, and have gaps underneath, or will be a rollercoaster of leaps and dips, with overly steep gradients, even on the most accurate DEM.
 
Last edited:
If you do not use the "Smooth Spline" tool (which flattens a 20m grid area on each side of the tracks), your track will be floating in spots, and have gaps underneath, or will be a rollercoaster of leaps and dips, with overly steep gradients, even on the most accurate DEM.

If he wants his route to be more realistic, then he will have to do what others do and be very careful and fill it in manually and avoid smooth spline , its tedious but the only way to avoid the artificial look one gets when the terrain is perfectly flat next to the track.This might be ok for a class one route but not for most others. I have a 60 mile dirt road that runs right next to the track on my route, next to a creek , if i used the smooth spline it would fill in the creek and cover cuts and the hills that come right down next to the track. I just have to use masses of spline points to do the road, which was never flat or graded, i do use the smooth tool on occasion but then go back in and move the surrounding terrain up or down.

BTW, I fail to see how his track will be a roller coaster if he doesn't use smooth spline , it wont move once you lock it into place. you can also use track with built in fill which will fill gaps under the track, or use track without ballast and sink it into fill under the track itself. This can also help get rid of the moire effect one can get on some track as you can cover it up a bit with ballast.
Is this time consuming ? , ya bet it is.
 
Yeah it's CN and CP class 1 mainlines (or former lines taken over by some shorts on the western side of the canal). It's located between Lake Erie & Lake Ontario with the Niagara River on the east side. I added some pictures of what I have done with Niagara Falls and the Niagara River in the screenshot section. So far this is fairly safe to do because I haven't altered the DEM in any way. But I'm getting close to the point where I'll need to focus on the track again like I was before the DEM.

The majority of the route is on mostly flat urban, suburban, and rural land. Where it gets interesting is the Niagara River, and the Niagara Escarpment. Now the escarpment and the Niagara gorge is a huge time saver using the DEM because I intend on including it as it's a scenic feature that is close enough to the tracks. both mainlines need to work their way up and down the escarpment (equivalent to the drop of Niagara Falls and Lake Ontario's former shore line in the last ice age). In fact the escarpment stays at or near the mainline between Niagara Falls and Hamilton (only going as far west as Grimsby for now which is about half way).

I think I'll just play it safe and keep the undo button handy. I have plenty of images my friend has taken at https://www.niagararails.com/ which gives me looks both ways at all crossings (similar to Google Earth but perhaps a bit more detail - not sure these days).

I'm abandoning the GPS elevation idea that I had because the GPS data is even more inaccurate than Google Earth. 15 zoom level virtual earth may also be inaccurate but the DEM is pretty on (thought it's only 70m NASA from what I understand).

This has kind of held my production of the route down. Before the DEM was gifted I had most of the southern CN main in place for operations, all the signals, and proper track alignments. Now as far as I can tell I can't transfer it so I have to start again. But I'm not sure of changing anything at this point considering I may be able to buy even more accurate data.

Lots of decisions to make.

Thanks for the advice, keep it coming.

Thanks

Sean
 
I forgot to mention some things:

Many years ago, I messed up my only original copy of my DEM by rushing in and pressing "Smooth Spline".
After 1 years work I realized that the grade from Altoona to Tunnelhill was not simply a constant 1.75% grade.
I had to press "Delete" route, and start all over again from scratch.

Now I have several unedited backup copies of the original DEM's, so I can start over if I should mess up.
Mine is a 4 track mainline so I use multitrack splines on straight away's, and single track on the curves.
Multitrack splines will not automatically raise, when a preceding single track grade is applied.
Eventually (after 2 years running through dirt mounds, testing out the grades) I replace all the straight away's multi track splines, with the same single track used in my curves.

So I set the grades to a preliminary height/gradient, and for upwards of a year or more, I test run right on through the humps and bumps, ignoring when I drive underground through terrain bumps.

After a year or so, when I am FINALLY satisfied, I clone my route again, and rename it slightly differently, such as: "DEM Gradients WIP 1 04 2020".

Then I make final adjustments on the cloned WIP DEM.

Rushing right in quickly with the "Smooth Spline" tool can ruin a DEM real fast, deforming the terrain.

Right now I have the gradients all the way from Rockville to Pitcairn pretty much set in stone.

The gradients are @ 0.05% to 0.25%, in some rare places 0.50% to 0.75%, except in the mountains, where it can be as great as 1.75%, in one place the downhill "Slide" is a 2.87% grade.

Of course I have thousands and thousands of cuts, and creek channels to rework, as the "Smooth Spline" tool has messed them all up royally, flattening a 40m wide swath of terrain.
 
Last edited:
...
2) If you make a cut or a mound in the landscape using the "Smoothing" tool; it can work well, but plan ahead, and use something VERY narrow - a walkway, path, etc - if you use the tracks or a road, the effect gets very wide. It also gets wider if you click more than once.
....

Ron,
Thanks for the tip about using a narrow spline. I'd never thought of that.
 
So I take it unlike Basemapz that uses an object, maps from TransDem paints the ground in a layer? So I take it you probably can't get as much detail on the ground paint as you would a base map object?

Now I said 30M NASA stuff before it's actually 70M if I read right which probably is the reason of even more distortion. I'm not sure where exactly to find better DEM through the province of Ontario or government of Canada sites. Ontario I had as little as 2M but wouldn't always reflect the true ground level because of tress, buildings, etc. The Canada website is a bit of a mess in my opinion or at least I can't get to what I'm looking for.

In some places it's difficult to ascertain how off the Visual Earth zoom level 15 is to the DEM. Which essentially brings me to nicer looking larger hills, but less than accurate placement than what I had with Basemapz (even that I had an error of about 1 to 2M per tile). I'm open to any suggestions on what I can use. I am considering a purchase of the program as well, as I'm kind of at a standstill. The person providing me the data is pretty busy, but I'm appreciative of any efforts or advice.

I just don't want to make a purchase of $50 CAD for 1 route and it turns out to be disappointing and less accurate for what is really important to me and that is the railway. The tracks for the most part are relatively flat in Niagara except for the escarpment.

Thanks

Sean
 
TransDEM does provide an object - UTM tiles - which typically sit slightly below the ground / grid. You control how detailed these tiles are by your choice of the source map, and how far you zoom in for detail. I believe there are four tiles per baseboard in Trainz. You also control how detailed the terrain grid is, by your choice of settings. However, the source of the DEM data will also determine the accuracy. Lots of variables!
 
TransDEM does provide an object - UTM tiles - which typically sit slightly below the ground / grid. You control how detailed these tiles are by your choice of the source map, and how far you zoom in for detail. I believe there are four tiles per baseboard in Trainz. You also control how detailed the terrain grid is, by your choice of settings. However, the source of the DEM data will also determine the accuracy. Lots of variables!

UTM tiles are optional. You can create the terrain with just the DEM data (and not having UTM tiles).
 
So before I was gifted this DEM (that has really left my development at a standstill) there was an argument that doing it manually would take much longer etc. What I'm seeing with DEM is automation but lots of extra work to utilize it.

Take the UTM's for example. Virtual Earth level 15. Now I can live with the fact that they aren't as crisp as Google Earth (I just need the reference spot). The issue I'm seeing based on the valleys and rivers is that the UTM is about 70M off to the left of the DEM and about 30 M off lower of the DEM. So basically every track that has been drawn (which is pretty accurate but obviously needs a bit of work and fine tuning, grades, etc.) is off with the map by that amount. Assuming that the DEM which is a 70M DEM is accurate enough.

The only way around this is to find a better UTM (not sure what is better than that in Canada) or start plotting in all the roads. Then decide on some landmarks based on elevation data to move everything a certain distance left and down to match the DEM.

If anyone has some suggestions I'd be willing to take them. I'm going on vacation in 24 hours and will be back Sunday night. Hopefully I have some answers because I'm lost.

Thanks

Sean
 
So before I was gifted this DEM (that has really left my development at a standstill) there was an argument that doing it manually would take much longer etc. What I'm seeing with DEM is automation but lots of extra work to utilize it.

Take the UTM's for example. Virtual Earth level 15. Now I can live with the fact that they aren't as crisp as Google Earth (I just need the reference spot). The issue I'm seeing based on the valleys and rivers is that the UTM is about 70M off to the left of the DEM and about 30 M off lower of the DEM. So basically every track that has been drawn (which is pretty accurate but obviously needs a bit of work and fine tuning, grades, etc.) is off with the map by that amount. Assuming that the DEM which is a 70M DEM is accurate enough.

The only way around this is to find a better UTM (not sure what is better than that in Canada) or start plotting in all the roads. Then decide on some landmarks based on elevation data to move everything a certain distance left and down to match the DEM.

If anyone has some suggestions I'd be willing to take them. I'm going on vacation in 24 hours and will be back Sunday night. Hopefully I have some answers because I'm lost.

Thanks

Sean

It sounds to me that this was done using very the old method in MicroDEM. Those old HOG maps are off by a substantial amount with roads and rails running up over hillsides, rivers not where they're supposed to be, and much else wrong.
 
Might I suggest the OP go to the Payware forum and ask Geophil in his TransDEM thread about what kind of detailed maps would be available for the area he is modeling. Geophil is VERY good about answering questions and he is easily the expert on what is available through his program. If it turns out that there are very detailed maps available that would certainly influence whether to purchase or not. ;)

William
 
Good idea. I hope he doesn't mind that someone is sharing a DEM for me out of boredom.

One last question if anyone can answer. Does the location of the DEM to the tiles always remain the same no matter how many times you generate?

This is an important one for me because if someone gets me started in say a 70M and I want to see what other ones like 2M look like for example, this is where I can really see a purchase of the software. Because then I'll have something that I can work on now, and upgrade later.

I take it there is no way of saving my old flat route? The tiles are off because obviously it wasn't aligned to any point on the earth other than what Google Earth said was north.

Thanks

Sean
 
Back
Top