16 replies to this topic

[Bjorn]

Please have a look at this blog post



Is it possible to do proportional symbol mapping accurately on a 3D globe, or should it be avoided? I'm not sure.

[Charles Syrett]

I suppose that, theoretically, you could do this accurately -- but how would people be able to read or understand such a graphic? By creating some sort of legend that would convey the effect of perspective on relative size? If your expected readership are mostly analytical types who would enjoy and understand such a graphic, then great! But my experience has always been that many (most?) people can barely understand even simple 2D thematic maps, never mind thematic maps that use perspective.

Eye candy -- yes! I'd love to pop one of those lemon bubbles into my mouth....

Charles Syrett
Map Graphics
http://www.mapgraphics.com

Is it possible to do proportional symbol mapping accurately on a 3D globe, or should it be avoided? I'm not sure.

[frax]

Björn, one problem you will have with doing this on a globe is the perspective. Are symbols in the distance smaller or not?

Also keep in mind that it is one degree harder for the viewer to assess the relative size of 3-dimensional symbols (cubes, spheres) compared to 2d, which again is harder to compare to one-dimensional (e.g. bars).

I am actually not too fond of cartograms with bars/cubes/spheres on top of a map, ordering and comparing the symbols is much harder compared to a chart.

[mika]

Is it possible to add some sort of colouring to the symbols (as per choropleth maps)? It would simply show the same value as the bubbles show and therefore it should make it easier to understand the symbols properly.

It's not always good to introduce redundancy to symbol's graphic, but I believe in this case it could help actually.

I thought of adding a label showing the value of the mapped parameter on the top of the bubble too, though it wouldn't be very subtle...

Dom

[Bjorn]

Thanks for your response so far - I'm still working on the issue.

Charles: I'm doing this for "non-analytical types". Maybe I shouldn't care too much about accuracy?

Frax: I agree with you. I'll do some experiments and post a few examples on my blog. Still, I think proportional 3D objects might be useful in some cases.

Mika: Coloring is an option. I'm using both colour and height for my prism maps, and I think it works better than only colour or height. A problem is that this is hard to achieve with 3D objects in KML, as you need to link to a different object for each colour (at the moment scale can be changed in KML, not colour). Adding a label showing value is an approach I've already added to my Thematic Mapping Engine.

[Charles Syrett]

I like the colour idea, especially for "non-analytical types". Then all you have to do is have one size of bubble and correct it for perspective, so that only colour defines the range. And oh, what eye candy that would be! You have smarties in Europe, right?

Charles Syrett
Map Graphics
http://www.mapgraphics.com

Thanks for your response so far - I'm still working on the issue.

Charles: I'm doing this for "non-analytical types". Maybe I shouldn't care too much about accuracy?

Frax: I agree with you. I'll do some experiments and post a few examples on my blog. Still, I think proportional 3D objects might be useful in some cases.

Mika: Coloring is an option. I'm using both colour and height for my prism maps, and I think it works better than only colour or height. A problem is that this is hard to achieve with 3D objects in KML, as you need to link to a different object for each colour (at the moment scale can be changed in KML, not colour). Adding a label showing value is an approach I've already added to my Thematic Mapping Engine.

[Bjorn]

Charles, I prefer M&M's Peanut! :-)

But I would prefer a choropleth/prism map than one-size coloured bubbles.

[mika]

I was actually thinking about duplicating symbology and putting it all together:

1. sphere / diagram size
2. colour
3. label

You may make it available as optional visualisation improvers, or perhaps make it a default option, so one has to tick them off if not needed.

The problem with 3d diagrams is that it's much more difficult to assess the exact value shown by them than when looking at the 2d diagrams, which are difficult to read for most people as well (see the attached comparison of circle and sphere radius for the same values).
J.J. Flannery did some psychological research on the perception of cartographic symbols and diagrams and eventually came up with a Flannery’s correction, an empirically derived exponent that has to be applied to 2d circular diagrams so they aren’t misunderstood and it’s possible to properly assess values and differences between them – which one is twice as big (shows two times more of the mapped parameter) than the other? Yep, try answering this question when looking at either 2d circle diagrams or 3d sphere diagrams… Spheres look pretty though

When I was at the uni they used to tell us – remember, whatever you do, try to explain it clearly in a legend so one can find it helpful… And the proposed legend for cartodiagrams looked like this:

 TestDiagrams.jpg   260.71K   57 downloads

Or like this:

 TestDiagrams1.jpg   207.46K   53 downloads

You can apply it to any shapes, but the main purpose of it is to give a reader a tool to assess presented values if needed.
I believe that legend of this type does the job well.

Dom

btw. today i tried to generate collada spheres with labels but got circle diagrams instead... A bug?

[Bjorn]

Hi mika,

Thanks for valuable feedback!

I will try do add a colour scale to the 3D objects later. The problem is performance as each coloured object has to be stored in a separate file. The benefit is that I can add a colour legend.

The labels could be on as default, it just looks bad when it's long values (population number etc.). I'll try that to.

I've been reading about the Flannery's work. It's interesting, but I've decided to only support mathematical scaling and not perceptual scaling at this point.

I like the legends. I can't see how I can use them since symbol size is changing when the user is zooming the map. The legend would have to be dynamic to work properly. But I would like to use the first legend for illustration purposes. What is the source?

Is "cartodiagram" an established term within cartography? A google search only returns 128 entries. Can't it easily be mixed up with a cartogram?

Bug: Are you sure that you selected the "Sphere" shape after selecting "Collada (3d)"?

[mika]

The labels could be on as default, it just looks bad when it's long values (population number etc.). I'll try that to.

You can always round the values and format them somehow. Anyway, that was just an idea, nothing obligatory

I like the legends. I can't see how I can use them since symbol size is changing when the user is zooming the map. The legend would have to be dynamic to work properly. But I would like to use the first legend for illustration purposes. What is the source?

Just a few minutes in DRAW this morning Feel free to use it if you want. I can send you a pdf as well so you can modify it.

Is "cartodiagram" an established term within cartography? A google search only returns 128 entries. Can't it easily be mixed up with a cartogram?

Well, perhaps it's because of the naming conventions in Polish and a bad translation on my side:
1. choropleth maps are in Polish called 'kartograms'
2. every diagram or other shape showing statistical values on a map is called in Polish 'kartodiagram' (carto-diagram I would say in English)
3. I thought cartogram in English mean a map where a reference area takes the size of a value showed (http://en.wikipedia.org/wiki/Cartogram)

Bug: Are you sure that you selected the "Sphere" shape after selecting "Collada (3d)"?

I am not since you mentioned it... It might be just me not paying enough attention to details...

[Bjorn]

Mika,

I would very much appreciate to use your legend, but it's something that bothers me:
I'm not using a full sphere, but a half sphere or a dome.

I guess the circle/dome ratio would be different from the circle/sphere ratio?

Bjorn

[Bjorn]

I see that the sphere/dome distincton can be confusing. I've now added a real collada sphere object to TME, so you can choose between cube, sphere and dome (half sphere).

[mika]

Well, let it be sphere, dome, cube, whatever one can imagine... The point is that assessing volumes is much more difficult than assessing areas. And it's already pretty difficult with areas.
The legend with dome would not be different in idea - you'd only use a different symbol.

So basically one could say the following:

1. 1d charts like bar charts show 'real' values scaled down to fit on a map - you can easily say which bar is bigger / longer and what is the relation = easy to assess a value
2. 2d diagrams use areas as mean of expression and therefore you're dealing with a square root of a showed value = relatively difficult to assess a value (remember Flannery's experiments)
3. 3d charts use volumes so you're showing a cube root of the value = difficult (very) to assess a value

See the chart with 'scales' of theese three different symbols and their relation to the source data.
 chart.jpg   31.17K   52 downloads

I like 3d diagrams because they look nice. But at the same time one has to always think what is the intended audience of the map and provide sufficient tools (legend) for a reader to 'enable' him to understand the map.

The other thing, already mentioned here, is making symbol's graphic redundant, so a few different parts of it show the same value - say colour, size, label, additional symbol inside a symbol and so on. When you look at small scale maps (say atlas maps - times comprehensive is a good example) redundacy in map symbols is pretty common, for example look at the city symbols and font size and type used to describe it. Colours aren't uncommon too.

dom

ps. I'll send you the legend example tomorrow morning (I am one hour ahead of you, so it will be morning )

[Bjorn]

Mika,

Thanks a lot for your informative feedback!

If I still want to use a 3D globe for my visualisations, I guess this would be the best alternative:



(KMZ)

The bar height is proportional to the statistical value. More information in this blog post.

I'm struggling a bit with the math (see this forum entry) , but when I read what you write about square root and cube root, it seems that I'm doing the right thing.

The chart showing scales with different symbols are interesting, but I'm not sure how to interpret it other than bars (1d) is substantially better than measuring area (2d) or volume (3d). Do you have a reference?

Nice if you could send me the legend image by email: bjorn@thematicmapping.org

Thanks again!

[mika]

Bjorn,

I never meant you shouldn't use a 3d globe for visualisation! Theese days it's the best way of propagating your data and ideas worldwide :-) Not to mention that GE, WorldWind or VE is easy to use for non techies. Try to convince someone to get sophisticated package to visualise some stats on a map...
It was all about the 3d symbols used to map stats...

And the chart... Well it was my idea of showing how symbol sizes (hight of bars, areas of circles or volumes od 3d objects, spheres in this case) change with values and how easy it is to differentiate between symbols. Simply, the bigger difference between two values (x axis) on the y axis the easier it is to assess the symbol properly.

The maths you struggle with - as you say I mentioned the same in this thread. if you're still unsure try calculating the collada (dome or sphere) radius using the equation I gave on the sample legend in a first step and then, when the max size is chosen, scale the symbols accordingly. I had no time to look at this but I might try next week.

And just one more thing... I know maths can be a pain in the neck, but you could try generating 3d (and 2d too) symbols using a logarythmic scale... Might sound scarry, but it isn't really. That could improve the symbol scaling though a legend is a must I think.

take care
dom