Ian Warren wrote:SO WHOS DATA IS CORRECT AND WHOS IS NOT , you may want to check the last screen to GOOGLE EARTH , you will notice the housing stagger down the side of the point , default autogen is well least to say not as accurate as photo placed autogen .
As a GIS guy this question comes up a lot- both when doing your own GIS work (why doesn't this match with what I just did?!) and explaining concepts of accuracy to others that may use your work/data/.
The short answer to Ians question could be anything from 'none of them are "correct"' to 'all of them are "correct" to some degree'- The main point is that the definition of 'correct' isn't a black or white determination and it depends on a lot of factors (i.e. what you are using the data for) and what you are comparing it to.
The long answer (well, kind of a short long answer....there are books devoted to this stuff)- Every piece of data has error, doesn't matter if you are measuring ball bearings, peoples political views or features on the earth (ie. Geography and GIS). There is no way to completely remove error (to achieve that would result in an alternate parallel reality! eek!) and trying to minimise it is a balancing act between cost (i.e. price to acquire better data via more accurate instruments, storing and distributing that higher density/quality data etc etc) and end use (would you guys want a 2m country wide mesh? Yes!...but not if it cost $5000 each). In this example, what is 'right' and what is 'wrong'....
I'll mention a few sources of error for 2 of the items of data that Ian has mentioned: Aerial imagery and Mesh
Aerial imagery-An aerial photo is taken of the earth through a camera lens which is a source of error that's moderately easy to remove. It is taken from a plane which probably isn't flying quite straight meaning the lense doesn't point straight down resulting in distortion. The surface of the earth has high points and low points which are a source of error since 'scale' is dependent on distance between the lens and the object, a hill is closer to the lens therefore objects on it will appear bigger compared to those in a valley.
Most of this error is removed when an aerial photo is turned into a 'ortho-photo': the image is corrected for lense distortion, the parts of the image near the edges of each frame are removed and the terrain error is 'subtracted' using a terrain model. It's good, but it isn't perfect- Pixels are generally within ~ 5 x the GSD (i.e. 1m imagery will usually be accurate to around 5m) but in some areas, especially on steep terrain, you''ll get stretching and warping to make it fit.
Ortho-imagery is the type of aerial imagery we use in FSX because they can simply be placed on the mesh fairly accurately and without too much fuss.
When it comes to FSX, you also need to tell the sim what pixels on the image are 'water' which is carried out by the person who converts the imagery for FSX- This is, you guessed it, another source of error! Around steep coastal cliffs especially, it can be hard to decipher which areas are water and which are land (not to mention that areas of land vs sea water change with the movement of the tides). To make matters worse, trying to make aerial imagery look good on steep cliffs is like trying to trying to take a good school class photo suspended above from the roof of the gymnasium- The ratio of pixels to area gets worse the steeper you get meaning a 1 pixel error in the water mask can extend a long way UP a cliff face.
Mesh (and water)- Generally speaking, a 'mesh' is a continuous surface created from discrete data- It doesn't make sense to capture every square metre of the land surface of New Zealand so instead the country was captured as contours and points. The land not covered by a contour or a point (which is the vast majority!) we don't
know about but can
infer things about. For example, a piece of dirt sitting half way between a sea level and a contour at 20m will be around 10m high, give or take. When it comes time to create a continuous surface from that data, the computer makes those kinds of reckonings: I've got a 20m contour here, a 0m contour over there so I'll this area with values between 0-20m....obviously it's a bit more complex than that but that's the basics of it (again, books have been written about terrain interpolation in GIS!). When it comes to FSX, water polygons override the mesh meaning that even IF the mesh says 'hey this area is land above sea level' the water polygon will say 'No it aint!'- Given that most orthophotos contain more information than the water polygons about where water is and where it isn't, this can lead to some issues in FSX as the lower quality data takes precedence over the higher quality.
So, to sum all it up- It's not as simple as saying 'X lines up with Y but not Z, therefore Z is incorrect'- It's simply the fact that error on datasets line up in a bad way sometimes. No data is immune to it
