Global LandClasses

[Guest luissa]

Hi,I recently came across the global database of land use available:http://www.gvm.jrc.it/glc2000/defaultGLC2000.htmHolger told me, in the mesh forum, that I should not expect much from it as considerable work is need to get good results. However I took it as a way to show how you can make landclass scenery with SBuilder starting with a bitmap whose colors represent "land uses". I download the map of Europe (binary folder) and opened it as a raw image in PSP. Then I changed the 18 grey shades to 18 RGB colors. All the information needed to calibrate this map was available. I just copied the North, South, West and East lat/lon border into SBuilder. The map defines 17 types of land: from metro areas to ice areas. I do not know if the same number applies for other areas of the world. I made a project file that you can open or import into SBuilder. Then you can assign the FS textures to each of the 17 types of defined land. James's PolyView available here at avsim) can be used in helping the selection the textures.Here is a picture:http://forums.avsim.net/user_files/76182.gifThe project (and map) is available at:http://www.ptsim.com/downloads/europeland.zipYou need SB version 1.06a available at:http://www.ptsim.com/downloads/sb106a.zipRegards, Luis

[rhumbaflappy 113]

Hi Luis.Landclass used by flight sim can be found here:http://lpdaac.usgs.gov/glcc/tabgeo_globe.aspand is the type of Global Ecosystems.Here is the legend:ftp://edcftp.cr.usgs.gov/pub/data/glcc/gl...tlon/oge2_0.legAs you can see, it is what Microsoft has used for their landclass data... even the names of the legend are familiar to us.Microsoft has enhanced this map concerning mountainous areas, as Holger wrote. They also have enhanced the urban designations. They have also extended the coastal areas, so that the default terrain in the sim always has a land type assigned ( or else the water classification would show near the coasts ).I believe that web site also offers continental breakdowns of the large download. Paintshop Pro version 8 can read the raw data.This landclass data is inferior to the default landclass in the sim, because MS has tweaked it.I haven't found a better source for global landclass. Justin Tyme has used available US land use data to create his US landclass... but he could testify as to the inacuaracy of that data as well, and the hours needed to fix it. He also used a lot of user feedback to tweak his data.The best source of land usage data would be accurate topographic maps, aerial photos, or eye witness for local areas. As you show, a painting can be made of the data. I have done this using a homemade palette with PSP on a 257x257 bitmap, then exporting the image as RAW. You could use any color scheme... it's the order of the palette that translates to the land usage number ( just like greyscale ).This should be a very good addition to your program. :)Dick

[Guest jimkeir]

Hi.I've had the JRC data for a while and I've been slowly working on converting it too. I work at the JRC and I've spoken to some of the people who have contributed to the dataset.In general the more populous areas are OK; Europe and Africa seem fine, China's OK and the American areas seem OK too. Some files, particularly the south asia one, are terrible though and the last time I looked, not all the world was available. Bits of Southeast Asia and northern europe/scandinavia were still missing, as are the extreme north and south. The global file has only 21 data values and so isn't really suitable.The data seems to match FS well, but has a few problems which you'll need to work round. These aren't necessarily data problems, but differences in the intention. There are a few data glitches too, however; I've found lakes on hilltops in Africa and a suspicious amount of ice the Middle East!First, coastlines and small islands don't match too well. You will need to extend the data outwards to avoid having missing landclass in FS, resulting in water squares around the coast. The dataset represents average landcover type, so if a particular 1Km square is mostly water it gets recorded as water.Second, some areas are *very* boring! They may be technically correct but look pretty dull; for example, most of North Italy is correctly marked as farmland but this looks terrible; you need to be able to change the source value for farmland into several different types, perhaps including several types of crops, occasional crops+town, maybe even a bit of forest here and there. Again, *on average* it's correct but the details get left out and that makes a big difference.Last, and possibly the biggest problem, is that in many of the files towns get largely ignored. They are there, but smaller towns disappear completely while the larger ones appear smaller than you would expect, and certainly smaller than the equivalent in the default scenery. Also, towns are marked as 'urban' - that is, no difference between a sleepy mountain hamlet and downtown New York. Naturally, I've got a little program to deal with most of this :) The only remaining problem is that of the cities and, as it happens, I think I found a potential fix for that last week. I've simply not had time to do much with it; I've been working on some other stuff and LWMViewer, and this has taken a back seat for a while now. However, if you want to see the JRC landclass data in action, you can try these files:http://library.avsim.net/download.php?DLID=47931 (East Africa)http://www.jimkeir.co.uk/FlightSim/Autogen/Falklands.zip... and I'll upload the Falklands area, and another in Africa as soon as the file library is accepting files again. For those who may have downloaded it before, the Falklands area has been recreated as a release-level package instead of the tech demo.I think the biggest problem for a project like this is testing! It's a little difficult to check the entire dataset to see if it matches the terrain, so there are *always* going to be quality problems.Cheers,Jim Keir.

[Guest flyfun]

Dear Dick;Landclass used by flight sim can be found here:http://lpdaac.usgs.gov/glcc/tabgeo_globe.aspand is the type of Global Ecosystems.Here is the legend:ftp://edcftp.cr.usgs.gov/pub/data/glcc/gl...tlon/oge2_0.legThanks for the intput!though a quick question - if you do not mind:site deliver .img files - what do you do next ? i mean - what program do you use to load and display img files - prior convertion?txmarc

[wsieffert 0]

Hi Luis,Why the need to revert back to sb106a vice sb201?W. Sieffert

[Guest Horst]

Hello W.Sieffert,Maybe I can help:the topic date of the thread is 16.05!You should use version 2.01!Do not use the download link for version 106a.Horst

[wsieffert 0]

:-lol !Thanks , Horst!!W. Sieffert

[rhumbaflappy 113]

Hi Marc.The IMG files are simply 8-bit raw data. They can be processed as any RAW file.But, this data is inferior to MS' landclass. MS has already extended the data into the water areas ( so shore areas are shown correctly ), and changed some values to match it's texture scheme. There's no reson to use this global data, or any other, as MS has done a good job with the global data.Tools exist to tweak small landclass areas of the world, and that's the best way. There's not much point to reworking the whole earth, if your data source is worse than the default FS9.Remember, you're dealing with 1km data, so don't expect miracles.Dick

[Guest flyfun]

Dear Dick;I appreciated very much your kind help!best wishes - Marc

[Guest GerrishGray]

Hi Jim (and everyone)Your remarks about towns as regards the LC data are correct for some parts of the world, i.e. towns tend to be represented too small because of the 1km averaging, but the reverse seems to be true for other parts of the world such as much of Europe and Africa, where the towns are often too large in the default LC mapping because very scattered development is getting counted as 'urban'. Small towns/villages of course nearly always get missed unless they just happen to be in the right place in relation to the grid.But there is no universal algorithm for making these corrections, it really does depend on the individual location. Burkhard Renk did a great deal of work on this when researching his excellent payware FSLandclass tool for FS2002 and came to the same conclusion. His approach was to allow the user to set some parameters that would make towns larger/smaller and woodland larger/smaller, plus quite a few other similar adjustments, for each LOD5 cell individually. But for all that, no mathematical modelling approach is ever going to turn the relatively crude worldwide data we have available into an accurate representation on a local basis.Fortunately the MS design for landclass, with its extra classes for various densities and types of urban/suburban, and the blending between adjoining classes, gives us quite a bit of scope for adjusting these problems on a local basis with tools like EZ-Landclass etc. Yes, I know that there are still some land usage types missing even from the MS set - such as industrial areas, and differentiation between old and new towns, to quote just a couple of glaring examples - but the design of the LC system is really well thought out and gives tremendous scope for local design and for future enhancement.We have been having an interesting off-topic discussion about localised landclass mapping under "Best place to place add-on scenery" ... (sorry, we should have started a new thread with the right subject heading really).Kind RegardsGerrish

[Guest arnaud.clere]

Hi,Regarding the GLC data, it has been used for France. It was quite accurate but still needed a lot of manul adjustments to:- better classify "artificial areas" that can represent towns or airports (very different FS landclass to use actually !)- extend the LC on the sea to accomodate with FS water/land separate classifications- fix moutain areas data that were really bad with altitude-based classificationAlso, I think the available data for France was far better than the one available for the rest of Europe. Be prepared for a lot of work...Regarding the urban areas classification. The GLC data was accurate for nearly every city with 10000+ inhabitants and was quite OK with 5000+ inhabitants cities. I had to randomly add small towns to get a realistic landscape in the countryside.Now, I am experimenting the use of geographic coordinates and population of every french town to replace the random towns with the real ones and it is very promising although it requires some elaborate programming to do so.Don't know about other places but for France, the freely available data allows to vastly improve the FS default LC. I guess it's even more true for the states.It's not as astounding as "photo-realistic" sceneries, but for large areas that's really nice to have a decent LC to start with.Just my .2 euros ;)