6 replies to this topic

[razornole]

Hello,

The projection was the cause of the linear artifacts, now my hillshades are coming out way too dark. There are hundreds of thousands of values between 0-150 but only thousands between 151-255. Why is this? In the attached example, the NW aspect pixels should be close to 255 but are coming out as mid-tone and the SE pixel are completely clogged.

Should I do something to my DEM before I generate a Hillshade, or can I just do something to the hillshade? Currently the DEM is 32 bit and the Hillshade is 16 bit. Changing the color ramp doesn't help because it is values that are wrong. The values should be consistent or evenly spread from 0-254.

Any thoughts?

kru

Attached Files

•  Dark_Hill.jpg   417.09K   36 downloads

[mika]

It looks like the vertical exaggeration of the dem is too high - if the values come from a 32bit tiff they don't match normal units as long as you don't match them appropriately. See what is the maximum & minimum value in your dem and adjust values so they correspond to the real world values or simply divide all the valueas by 10 or so.
Attached is a quick example of srtm3 data - all the dem cells were multiplied by 10.
Hope this helps :-)
 test.jpg   6.07K   33 downloads

[Sergio Huykman]

Hello,

The projection was the cause of the linear artifacts, now my hillshades are coming out way too dark. There are hundreds of thousands of values between 0-150 but only thousands between 151-255. Why is this? In the attached example, the NW aspect pixels should be close to 255 but are coming out as mid-tone and the SE pixel are completely clogged.

Should I do something to my DEM before I generate a Hillshade, or can I just do something to the hillshade? Currently the DEM is 32 bit and the Hillshade is 16 bit. Changing the color ramp doesn't help because it is values that are wrong. The values should be consistent or evenly spread from 0-254.

Any thoughts?

kru

Hello, do you tray with a minor vertical scale? If default is 1, then put 0,0001. If something happend see some tutorial page in Esri page to find the explanation. Sorry about my english, I'm from Argentina.

Good luck
Sergio

[aug_aug]

These are handy, this might be the problem but I'm not sure which dataset you're using above:

If the input raster is in a spherical coordinate system, such as decimal degrees, the results from Hillshade may look peculiar. This is due to the difference in measure between the horizontal ground units and the elevation z units. Since the length of a degree of longitude changes with latitude, you will need to specify an appropriate z-factor for that latitude. If your x,y units are decimal degrees and your z units are meters, some appropriate z-factors for particular latitudes are:

Latitude Z-factor
0 0.00000898
10 0.00000912
20 0.00000956
30 0.00001036
40 0.00001171
50 0.00001395
60 0.00001792
70 0.00002619
80 0.00005156

[razornole]

Thanks for the replies,

I thought that this was going to be a simple problem (which it may be to most but not me).

Everything that I did was default with ArcMap 9.3 with spatial analyst. The DEM is from USGS NED 10m. I mosaiced several tiles and then ran the Hillshade module.

The values of the Z are metric and they are real world values (203m - 1952). The lat/long is in decimal degrees ~ Lat 35, Long -84. The z factor is 1.

At least now I know some key words to search for in ESRI help.

Thanks for your help,
kru

[razornole]

Hey Michael,

I tried your method and it worked like a charm (see attached). However, I pulled a random z-factor of .00001075 because my lat was in between the figures you gave. Is this an exact science (i.e. is there a formula/algorithm to get the z factor) or does one just want to try to get close?

Thanks,
kru

Attached Files

•  Better_Hill.jpg   482.47K   35 downloads

[Mike Boruta]

It is ok to approximate the z-factor. See here:

http://blogs.esri.co...-correctly.aspx

But it might be preferable to just project the DEM before making the hillshade. Then you can leave the z-factor at the default of 1. Just make sure to scroll down and change the resampling method from nearest neighbor to bilinear (or you'll get those grid-like artifacts)