Atomic Systems

Oh what to do! I am running some track through about six miles of some very hilly countryside. I am using Nobkins RWDecal and of course the landscape was done by importing from the DEM files. The Google decals are quite high detail so I can place track fairly accurately. So far so good!

To get the gradients correct I am using Ian Allens "BR Mainline GRADIENT PROFILES: The Age of Steam" and have measured mileage along the track almost to the meter in order to ensure the accuracy of the route ... a little bit out in one place can throw the whole route out miles from the scene of the discrepancy.

Here's the problem and I'm sure others have run into the same thing: the rail gradients do not match the terrain. The DEM elevation points are very accurate, I believe Google mapping is fairly accurate so it must be something that the RW grid system does with plotting the DEM. In one location I noticed a stream running alongside a hill instead of in the valley, also a three hundred foot station platform that dropped fifteen feat along its length in what should be level terrain. Starting from a known height I laid the track from A to B using the correct gradients along the way. On fairly level land or gentle slopes there were no problems, even small embankments and cuttings were all in the correct places but its when I started getting into the hills the trouble started. Track started plunging up to four meters underground where it should be on the surface of an inclined valley. It got so bad that eventually, strictly adhering to the gradient profiles, about sixty percent of the route was high in the air or underground. This would require cuttings, embankments or tunnels where there are none. On Google street view and ordnance maps the tracks are all on the surface with no cuttings or embankments. The alternative would be to lay the tracks along the surface paying attention to where there should be cuttings and embankments and make it visually correct but the gradients would be pure fiction and detract from the driving experience. The last choice would be to keep the gradients and modify the landscape over a large area which is very hard to do in the hills and valleys and would once more lead to a fictional route.

I would really like to know what developers of mountain routes do in such situations. I'm sure they must have the same problems and where they make their priorities when it comes to gradient or terrain?

Any help would be appreciated.

Mike.

This is a difficult one and I don't think there is a single answer.

I would normally lay the entire track first based only on gradient profiles and let it fly through the air or through hills and then check progress at various key points where it is known that the track sits on the terrain and is at the correct elevation. This would indicate that despite what might seem wrong coming up to that check point, the gradients are correct. It is then that you would need to mould the terrain to match the track and perhaps make adjustments where the result was not as expected. I think DEM information can be incorrect in places but the Google overlay should be completely correct and as long as the track follows the overlay then you cannot go far wrong.

There is clearly something wrong if you start at one point and do not end up in the correct location with your next check point using the gradient profile alone, but at least you know where to look for the error.

Check that you are using the correct measuring units from the gradient profiles (yards, feet, metres, etc). I was completely puzzled by the Bluebell Railway gradient profile which is measured in miles and chains

I hope that helps.

-

I would agree with all of Geoff's comments.
I think you will find the gradients are accurate (assuming any conversions are right) but accuracy of DEM heights are variable - especially within the UK.
I know for a fact that around a seashore that I am contemplating a route build, the land was way out but as it was the Datum point, I lowered the sea-level to match the embankment height that I needed!!

However, things shouldn't be too far out so you may well have a error somewhere in the conversions. But the final build is usually a compromise on gradient & DEM.
I would say the gradient is the priority so that it drives properly and massage the terrain to match.

Good luck

Steve

Hi,

I did some research on a short 11 mile route that I'm building and I found that GE couldn't always be trusted - the main problem area was caused by buildings which seem to artificially raise the terrain around them by quite a distance. Research on gradients showed errors and discrepancies between LNWR Gradient Profiles and BR Gradient profiles. Any areas with mining will see gradient changes over the years due to subsidence. At one end of the route the gradients didn't even come close to matching the GE terrain. I visited the area and saw that old colliery slag heaps had been levelled and land built up and large warehouses erected. This had caused radar imaged GE terrain to be significantly above the original level. I did a graph to compare the GE terrain and the LNWR Gradient Profile.........



The terrain spike at 2.75 miles is due to a large industrial building and the difference at Coalville is as described above.Differences at 7.5 miles are probably due to buildings but not sure at present - needs more research

So it seems our routes should be divided up , known correct heights identified - for example where there are no embankments or cuttings - building heights should be allowed for - the ground needs to be modified in places and gradients might need some modification where there are obvious anomalies - there is no hard and fast answer - a little artistic thinking is required, but I think the main changes should within reason be to the terrain. Very large anomalies are probably due to calculation errors and wrong gradient info multiplied over distance - Perhaps route heights should be calculated in both directions from start to finish and vice-versa from known and accurate heights to try and show any obvious errors - I think very long routes would need to be divided into sections and this process applied ?

I checked the GE terrain against the RW terrain and I found it to be very similar along this short route.

Cheers
Paul

My experience is that neither the DEM data nor the gradient profiles typically available are very accurate. My approach now is to put the gradient profile into a spreadsheet, establish benchmark track heights along the route (at level sections or from OS maps), then tweak the gradient lengths (rather than gradient values) to match. Even so some manual terrain adjustment is usually required.

I found something interesting. I was at a loss as the end of the route was way up in the sky and the beginning was underground. I had done a save and was just casually fiddling with the route and decided to tilt the whole route to see what would happen. I did this by using the gradient tool and making sure that all the red triangles along the route had corresponding red triangles on the up or down track next to it (tracks become unaligned when tilted if this isn't done). Would you believe that the route almost matched the contours and the gradients had tilted with the route and remained the same. Maybe the gradient is relative to the tracks around it and not to the terrain. I'm leaving things as they are now and will tweak the landscape. Hopefully the end result should be close to the real thing.

Thanks for your help guys and I will probably be back again later for more tricks and tips

Mike.

The satellite view from GE can also be out - sometimes by 100m.

Also, from experience down here (which used to closely follow UK practice), gradient diagrams are somewhat generalised and there can be some small discrepancies.

I am in agreement that gradient profiles have small discrepancies. When the track is inclining at 1:70 then declines at 1:200 for example, it causes quite a noticeably sharp change in the tracks that if in real life would be derailing trains. The gradient has to be changed into a smooth "hump" over at least 60 feet and that change is not reflected in the gradient profiles.

Mike

Keep in mind that the accuracy (or lack thereof) lies mostly with the DEM, which, depending on the resolution, can be as much as 180m out for 90m (nearly 600'). If using 10m DEM, then the accuracy is + or - 10m for any given point, and 30m would be + or - 30m. LIDAR data is the most accurate (+ or - 3m), but is not available for many areas (at least not free - especially outside the US). Google uses SRTM (Shuttle Radar Topography Mission) data, which for the UK is 90m (30m for the US only), which is probably the reason for the discrepancy. Railworks supports up to 8m, so is you can find a good free source of 10m DEM, then jump on it. 30m is a decent compromise, but 90m (as in Google Earth) looks absolutely horrid.

As for gradient (profile) charts, accuracy in the States is typically to the nearest foot (from sea level) for each 1/2 or 1 mile. I can't imagine the UK being any less so.

To clarify, the DEM data resolution refers to the horizontal distance between sampling points rather than the accuracy of height data. The 50m Ordnance Survey Panorama Landform data most of us use here for UK routes provides a height value at every intersection on a 50m grid, which is then interpolated by TS to shape the 12m terrain grid. Thus there is quite a bit of smoothing out of sharp dips and ridges. in addition the height data itself is subject to measuring error which I believe Landform quotes as plus or minus 3m.

There can be a big difference between 90m and 50m resolution in hilly areas as you can see from my screenshots in this thread:

http://forums.uktrainsim.com/viewtopic. ... t=landform

Noted. Thanks for the correction.

I found the best way was to mark each mile with a separate marker and carfully adjust the graidient changes to them, this has worled all the way from Amber Gate to Gowhole on the midland main line.
The problem I found is the dem isent as acurate as it seems, remeber there in 30 degree arcs and this does lead to some jumps

one day we will I hope be able to use LIDAR but thats in the stars at the moment

This might be of interest to you -

http://forums.uktrainsim.com/viewtopic. ... &sk=t&sd=a

-