5 replies to this topic

[burwelbo]

I am trying to figure out what the ratio (if any) is for mapping raster data at certain scales. What I want to do is to produce a hillshade and add texture to it. I have been experimenting and I have noticed that you tend to loose some of the data when plotting or on screen due to what I assume is the resolution of the data. So the question is there an optimum pixel size ratio to map scale? For example, If I am making a regional map at 1:500000, should I be generalizing the results of my analysis to say a pixel size of 250 m or larger. Has anyone else had this trouble and is it just trial and error?

Bruce

[Rob]

Bruce,

that's a very interesting question, to which I don't think there's an 'optimal' answer.

I've always gone trail and error, because I find each project that requires a shaded relief needs one that fits nicely into the overall hierarchy of the project, and depending on the map's theme that could be any where from near the bottom to extremely obvious. I also like to generalize/smooth the data as a scale dependent function in addition to culling data out with the grid spacing.

Obviously, at some point the data will start to break up or contain more data than necessary, so you could find a ratio of where that happens. And if anyone does, I'd be curious to know what it is.

rob

[pghardy]

The classic paper on this matter is by Waldo Tobler in "Measuring Spatial Resolution". Proceedings, International Workshop on Geographical Information Systems, Beijing, 25-28 May 1987

I suppose that that is nearly 20 years ago, so 90% of people in the industry will not have encountered it!

The introduction starts "Spatial resolution is defined as the content of the geometric domain of observation divided by the number of observations, all raised to the power one over the spatial dimension. The size of the smallest detectable feature is twice that of the resolution. The usefulness of a Geographic Information System is constrained by its spatial resolution."

Table 1 of this article gives:
Scale Resolution Detection
1:10,000 5m 10m
1:50,000 25m 50m
1:100,000 50m 100m
1:10,000 125m 250m
1:10,000 500m 1000m

Toblers first rule of geography is: "Divide the denominator of the map scale by 1000 to get the detectable size in meters. The resolution is half of this amount".


I have a analogue copy that I could email you if you contact me directly.

There are other tables and articles available on the web relating scale, accuracy and resolution - as an example see
http://www.profc.ude...l2/cp1/t1-2.gif

Regards,

Paul Hardy
(phardy@esri.com)

[Martin Gamache]

In addition to Tobler's numbers I have seen the following rule of thumb when working with satellite imagery:

5 m pixels (IRS) good for compiling maps up to 1:50,000
10 m pixel (SPOT) 1:100,000
15m (ASTER, Landsat PAN) 1:150,000
30m (Landsat TM) 1:300,000

This is for extracting plannimetric data such as roads, trails, from said imagery The equation is probably different for deriving landcover...

I'll dig up the exact reference when I get into the office this morning.

[Hans van der Maarel]

For a number of projects (roadmaps) I've displayed aerial photos underneath polygons to give a bit of visual appeal to areas that otherwise would have one solid color. Especially in rural areas this causes a nice effect.

 hoenderloo.jpg   170.76K   138 downloads

The resolution I use for those photos is 2m, scales are between 1:10.000 and 1:15.000, files meant for both on-screen and print/press use.

[Martin Gamache]

Dr. Karsten Jacobsen from Hannover University has numerous papers on the topic:

http://www.ipi.uni-h...#Publikationen:

In particular;

Comparison of Information Contents of High Resolution Space Images

Comparison of High Resolution Mapping from Space