Is this data good enough?

The UK Ordnance Survey are giving away free their opendata range for the entire UK, which includes individual building levels data, detailed coastline, important buildings, water areas, woodland and some others.

The licence info seems to be really relaxed :)

I downloaded the OS SW tile to look at some of the data in QGIS - reproduced below. I've not tried to use ScenProc at all yet, will this data be good enough for some decent results do you think?

Thanks K
licence info.jpg SW data zoom.jpg QGIS data example.jpg
Last edited:
Wow, Kevin, this looks like the "bees knees"! Thanks for sharing. It's about they did something like it given all the prior public investment in the OS.
I wish we'd had something like it for work I was doing 30 years ago :)yikes: oh boy, 30 years, before what they call GIS today. It was before I left the country. We hand digitized the UK (the transport infrastructure - my background is logistics planning and related systems). What a chore that was! Got to know the good old 1-inch maps well though! They were just bringing out some digital data - on the old massive tapes, but wanted a fortune and we were already done with Britain before they covered even a tiny part of it.

I'd say this is perfect for scenery work - whatever techniques/technology you use. Now I just need something to match it for North America... :wizard:

Haven't thoroughly checked it out yet but did you notice whether there's a lot of data on roads etc.? :rolleyes:



Staff member
FSDevConf team
Resource contributor
Hi Kevin,

Looks great indeed. I would also check the documentation on the attributes. If the attributes are also detailed (like vegetation types) making autogen is much easier.
unfortunately that info doesnt come with the free stuff Arno. I do however from another source have 1km sq multiple tree species data for the whole uk showing where each species can be it should be possible to tie that into vegetation groupings which could be applied to localised areas?


Staff member
FSDevConf team
Resource contributor
Should be possible to combine that indeed.

But even without documentation you might be able to find interesting stuff in the attributes just by checking the data.
The Trimble OSM data for the UK seems to be quite interesting as well - thanks to GaryGB for pointing that out in a recent thread...
Power line data seems to drill down to pylon and pole locations for example. Should be interesting in conjunction with other data. Thanks Gary!

Trimble Cornwall.jpg
look at you starting to have fun with data :)
there's lots of data in that set; without being able to decipher it its pointless,
you can probably classify a small area for your project with this; such a shame they removed that information (have no doubt someone has the key for this data!)
Delving into some of the data attributes I think a lot of it _could_ be useful. The real nugget would be complete Uk land use and building data, but that is all licenced stuff here with large pricetags attached! :p Still, there's enough out there to make a reasonable stab at some UK 3D autogen, when:
1) I have the time to work out how to pre-process it for maximum value; and
2) I work out how to process that data in Scenproc properly; and
3) My wife lets me :D
No - Ive been reading the scenproc manual :) Much easier to understand when there's simple worked examples...
Can't comment on no 3 though :tapedshut
From Trimble OSM data above, I'm seeing some nice possibilities:

1. Bus routes - possibilities of creating living world traffic types in the correct places / Thousands of bus stops identified - for autogen placement
2. Thousands of instances of fence and hedge line data - for additional 3D autogen
3. Hundreds of waterway instances - could be used with FSET and SBX for highly accurate watermasks
4. Public transport points - accurate placement of bus stops
5. Hundreds of shops identified - for autogen building clsssification
6. Hundreds of campsites and hotels identified - for autogen building classification
7. Weirs sluices and waterfalls - for XML models?
8. Thousands of amenities identified ie pubs, telephones, postboxes, places of worship - for autogen placement
9. Hundreds of gates and cattlegrids identified- for autogen placement
10. Hundreds of historic points ie derelict chimneys, standing stones, abandoned mineworkings, wart memorials identified - for autogen placement
11. A hundred playgrounds identified - for autogen placement
12. Communications towers and lighthouses identified

Building classification should be made easier through use of the data above to select out various building types, for example, all buildings identified could call a specific autogen class, eg pubs, buildings outside generalised built up areas could be classed as farms, key buildings identified as schools etc. I'm building up a small but accurate database of land use data which will help determine building type :) I was also imagining customising the building textures to actual localised building imagery for each 5km sq area? - for Cornwall that would be 100 variations alone...

Autogen vegetation could be identified through natural world Geofabrik and OS woodland data, supplemented by scenproc's detection facility?

Motorway, A, B and minor road data can determine different road traffic densities?
there you go mate; think outside the box :)
did you make sure your waterlines are accurate (compare against live imagery)
it is not completely useless; without road classification and buildings throughout it is going to be a task of classifying data first,
you can cross reference with other data-sets and import information; thats data mining for you ;)

yes you can detect and build your db on your own; you can also use road width to try and classify a generic type,
it may take you a long time; but you will eventually have the data where it makes more sense,
Road classification is there already Chris. :)

I think I have enough land use data from open sources now to make a reasonable stab at industrial areas as opposed to residential ones, especially when combined with building footprint size.

Remember I'm not attempting to get EVERYTHING 100% accurate, just make a better job of it than having default autogen across the board. :p

One of the larger obstacles for me will be modelling all the specific bits and bobs I want to include in new autogen classes, as modelling is not my strong point. :(

It will also take quite a bit (LOT!) of time for me to do the necessary texture preparation for the many autogen texture sheets variants I'd like to do. I really want to do that because I can see that would have an immediate diverse visual impact which would aid immersion considerably.

I've played with the Scenproc vegetation detection stuff quickly and it seems to be pretty easy, I've now downloaded 60cm tile data through FSET for a test area of Cornwall so I want to run that through to see how much it adds to the vegetation data I already have.

Not checked waterlines yet, for another day/month that one! :p

QGIS Cornwall current data.jpg
Potentially of benefit to those seeking UK data:
National Forest Inventory outputs -
English indices of deprivation 2015 - ; allied with..
Lower Level Super Output Areas -
County Councils - ie Worcestershire - (includes things like exact streetlight and grit bin locations!) Have to be looked at on a per council basis and although freely available to view online this data isn't available to download freely.
Corine land cover 2012 for the UK -