FSXA Population density data ranges

[FlightGirl]

In a project for a region I'm working on I'm in the stage of put in the last details and I've got some quite accurate population density data to be resampled to fsx. (Yes, I know it is just he icing on the cake... but... anyway... some of us think that's important too...)

The value for FSX is a percentage value between 0% and 100%. I've searched a bit but can't find any reccomendations for a common or reccomended way to classify the density for use in fsx.

Are there any reccomended ranges for the size or range-values to use for the PopulationDensity?​

I have data for persons per square km, and also the number of people living there (and also even down through 500m, 250m and below, but I guess the 1000m is the best LOD/accuracy match).

The data also contains a palette index from 0-5 grouping the sizes in some uneven ranges like

0-4, 5-19, 20-199, 200-499, 500-5000, and 5000+.​

Is this a common standard? And is this maybe a good way of grouping?

I would like to avoid having extreme traffic or boat activity in a small village, but I'd like to have some boats on lakes where there's small towns.. -And if I used the full range and gave it a percentage for the full range between 0-5000 this would end up with the most of the region below 2% even for local towns. And usually such places may have some leasure boat traffic...

My intentional idea was to group it based on some scale like this:
- below 200 / km2: the actual percentage, but from 0% to 70%
- below 500 / km2: for the scale between 70% to 80%
- below 5000 / km2: for the scale between 80% to 90%
- above 5000 / km2: for the scale between 90% to 99%

I see in the sdk that the WaterClass has some depth-ranges like 0-5m, 5-x ... and up to 80m and deeper, where it stops. This may indicate that such an accuracy may be good enough for the population density(?)

And looking on the populationdensity.bgl in fsx it ssems like they use something similar too. But there's no more information on this.

Is there someone out there that has made some reccomendations in a forum post or a whitepaper?

Has some of the products like Orbx or Scenery-tech established a standard we could follow to ensure consitensy? And so user don't have to adjust their sliders depending on the product they have running.

Do some of you have any thoughts on this matter?
(Or even the solution ;-) ...or just want to discuss the concept?)​

PS! Forgive me for my long post and if this is the wrong forum. Maybe it's more of a Landclass or LivingWorld... but that seems to be focused on AI, ...and this is related to the resample sdk. So I was hoping this was an appropriate plase to ask.

 

[arno]

Hi,

Very interesting questions, you put them in the right forum.

I do also have some data here that gives me population density data in squares of 500 x 500 m, but I never tried to put it through resample. I should have a look at it .

I think there is no standard on how to represent such data. My data just shows the population density as a number if I remember correctly, no grouping in categories or so. But I can also imagine that some categories are defined and used.

The sdk indeed is not very specify on how many people per square meter belong to a value of 100. But maybe it doesn't matter that much either, since it influences things like traffic and you won't see the difference between density numbers that are close together.

My approach would be to check the density in a big city and then use that figure as 100. Then you can scale the rest against that.

I'll see if I can have a run with my own data and see how it works out

 

[rhumbaflappy]

The density is in a range of 0 to 100... I believe this is a percentage between no population, and "maximum" population density. It would be subjective as to what maximum represents as humans per kilometer squared.

You could paint 8-bit greyscale bitmap to represent the population percentages, and resample that way...

Dick

 

[Operator]

The value for FSX is a percentage value between 0% and 100%.

Actually, 0...255%, it depends of color of BMP. Here is an example, look at the value:

 

[rhumbaflappy]

0-255 certainly makes more sense as a range. It could be the SDK is in error... or it could be numbers over 100 are clipped or ignored. It's hard to test, as this controls types of traffic (?).

Dick

 

[Operator]

I tried to compare 100 and 255 and found that it works. With value 255 there were much more boats and car traffic than with value 100.

 

[FlightGirl]

Thank you for the feedback, and it's good to see that this may be interesting for others too.

And, yes Dick, it may seem that the "answer" is close to what many are saying: we have to first try it based on general density (with the biggest city as 100). And then run it throgh some tests and adjust it.

And after all this is denity per square km (or some smaller units) so they should be comparable in some sort. But the problem remains with the huge difference between large cities and smaller towns. A lage city may have more than 4500 persons per square km (and (edit) yes, I noticed some over 16 000 in the middle of Oslo), but a small regional town may have as low as 200, and in an average my country has 7(!) persons per square km. And I still want some leisure boats outside that regional town too. Although I can see that also the small tows has at least one square of high-density in the middle of it's centre...

And it would be interesting to see your findings Arno, since we represent maybe two of the most extreme opposites as to population density in europe. Norway has one big city, a lot of small-to-medium towns/cities, but the vast majority of the country is sparsely populated, but the Netherlands has several large cities and a much higher general density. So if we can find something that works with both, then it should work for many areas in between

I've seen that the fsx-file has added some higher density along the major highways. Thinking about it: that may be the way to go: to add up a couple of sources....

But I'm still wondering what the consensus should be - as I don't want to use a much different density in my area compared to other. That would couse the user to adjust the density-sliders when moving around the word in comparable regions (or suffer from a possible drop or increase i fps).

@Vitaly: Even if it may be possible to use the values up to 255 I'm reluctant to step over the boundary of the SDK. Maybe they should be reserved for "extra large densities" like Mexico City or Tokyo. Maybe you also point out an important issue that there are some regions staying within the percent 0-100 and some others using a full byte-scale of 0-255.... Or maybe Dick is right and that it just crops it?

I'm looking forward to see more of what your finding are, and I'll post my findings too.

Christine

 

[FlightGirl]

I just saw your answer Vitaly. It's good to know that it works. But I'm still wondering if we should... I'm looking for a general representation that gives th natural amount of traffic and leisure boats... And without filling it with too much (unless the user turns up the sliders)....

 

[iangbusa]

You could also try "normalizing" your complete range of data as an index 0-100 (0-255) with your maximum population density at 100., using those values as the value set in the bitmap file to be resampled.

 

[arno]

Interesting idea to also add the highways to the population data. That should be a good way to ensure there is enough traffic on main roads linking important cities.

 

[FlightGirl]

Yes, both the normalizing and addition of several layers are spinning around in my head right now...
The idea:

1. Use the base pop.density data (and yes; normalized in some way. But maybe not if we combine it)
2. Use the road-data (maybe with a buffer of 500 meters, and only the motorways with a value of 80% and primary roads as 50%)
3. Add some other like for instance a user-adjustment layer for popular recreational or tourist-like lakes

And then just use a simple weight for each layer: for instance: layer 1 as 70%, level 2 as 20% and the last on 10%.

The roads will then make sure there's both intra-city travelling and also most of small-towns will be located close to major roads...

It sounds complicated... But it's an interesting idea....

 

[arno]

Hi,

Yes, I have similar ideas spinning around my head now .

My population density data is in vector format, so I will first rasterize it. Then I will also try to normalize it in some way. Not sure yet what value to use as a cap and it I will use a linear scaling or not.

Then I will burn the road data into the raster as well. I think I will multiply the population density based on the highway in some way. Because highways in dense areas have more traffic than highways in low density data. Need to experiment a bit with this.

Hopefully I find some time tonight to play with this,

 

[FlightGirl]

It seems that the cap used in many diagrams are in the range from 1500 to 5000. Maybe that's more than enough ro indicate that this is the highest density city.

And just as you say: this will be helped with the roads as cities has many and large roads. And those local towns that got a low "score" since we went from 1-5000 will regain it when they are close to a major highway (as they most certainly are).

...this is growing on me. We may be on to something.

 

[arno]

Hi,

I did some more thinking. Instead of adding or multiplying the roads, I think I'll try to use the maximum population density in the area around the road. That way the density of cities is kind of spread out along the roads.

But maybe I need to check what else besides traffic is influenced the the density BGL. Because else it might be easier to just use a lower resolution data, that will blur things out automatically

 

[FlightGirl]

Yes, it needs some thinking because it isn't that straight-forward.

I tried the base density data but that gave too much difference between the high and low density area and with too much of the rural (but still populated areas) below 2-3%. So maybe my initial thought with classifying 0-4, 5-20, 20-150 and so on will be a better base...

So the conclusion may be that the data got too high resolution. Or maybe some kind of a heath-map will help.

 

[arno]

Hi,

I had a look at the default population density data. It seems for the Netherlands Microsoft already decided that the minimum value is always around 6, even for the coastal sea area and bigger lakes. On land I also don't see much lower values.

It seems the really low values are only for the Sahara, Greenland, that kind of areas.

 

[FlightGirl]

It is about the same for my country. Then it might be possible to think in terms of taking that (or an country-average) as a base (even out in the sea). And then add the roads (but with an additive matter). Let's say that a single road within a cell is 5%, then a single rural road will end up with 6% + 5% = 12% and in a more dense area there's maybe 10 roads in a cell(giving 56) and in a city there may be more than 50 roads per cell it comes close to a 100.

...but then, what should I do with my fine grained data I was so proud of

 

[FlightGirl]

Well, I've done some experiments on this but that lead me back to reading a bit of representing statistics spatiallly. And this lead me back to see that the original question(!)

It seems to me that this is a common problem in how to best represent distribution of statistical data. And of couse: so it meet the requirements of the task. My initial goal was:

1. Making it comparable to the rest of the FS-world so the user could jump between different regions without having to adjust the sliders.
2. Follow a common consensus from scenery developers.
3. To have the correct representation of traffic and leisure boats.
   1. This includes both increasing (a tiny bit) around some centres
   2. But also decreasing (or even remomal) of boats or cars on vast non-inhabitated areas.

To try to summarize the work and discussion [so far] a bit:

One path, is to just take the raw data: But for this task it leads to some problems:

1. It's got too high distribution of data. We will end up with a too high concentration of areas in the 0-3. And for our purpose there's a huge difference between for instance 0 and 4.
2. It's too high detail/granualirity that leaves out areas connecting settlements or lakes/shores

We tried to experiment a bit with this with combining it with other data. For instance:

1. Roads can help on pt. 2 above
2. The number of roads can help on pt. 1 above.

But still the problem exists: There's still too hight distribution between the high and low leaving a need to diffrensiate in the lower end of the scale.

Other advices is to decrease the details. Make the grid larger so that more adjecant land or ocean will be covered.

Reading about it, I see that it's a common problem that has a lot of different theories. I have a couple of books and also found a few articles on the net. But I want to restrain myself rom going too academic....

But this leads me to think that there's a reason why the original data contained a legend gouping like:

0-4,
5-19,
20-199,
200-499,
500-5000,
5000+.​

This is after all made by some very skilled statistical minds. Of course they have not this task in mind, but they have probably some degree of experience and knowledge of the data (like the statistical distribution).

I can't use all my time on this since this isn't the most important issue in scenery development. And after all the defaults from Aces isn't to bad (actually quite good). So it may progress a bit slow, but here's what I'm thinking about doing:

1. Assign level according to the groups above.
   1. For instance 10% or 20% intervals.
   2. Maybe downsampling to for instance 2 square km blocks
2. Add a roads-layer.
   1. And for each road I add either 1% or 2% to the interval above (i.e. for a vaste unhabited area a road will add just a few percent, for a small town with 20% from the base, and with a three major road passing through will then get 26%).

Feel free to comment some more or share your findings or reccomendations.

C

 

[rhumbaflappy]

It's too bad we don't know the source of Microsoft's data. Landclass was originally the Olson classification. I don't know what source was used for waterclass.

Dick

 

[FlightGirl]

Yes, you make an important point there.
I found a few web-sources for world-wide population, water depth, plankton-concentration and so on. And I'll collect them later and post it.
But the problem we still have to figure out is what this should translate to in the fs-world....
But without it we wouldn't have all this to discuss

Christine