Strange hillshade edge effects around pixels

423
4
Jump to solution
07-07-2021 01:50 AM
David_Brooks
MVP Regular Contributor

Can I get some advice on what im seeing here?

Below is a continuous raster with hillshade applied.

On the right is a 5m terrain dataset

On the left was a 50m terrain dataset that's resampled to 5m (with nearest neighbour interpolation) and was mosaiced with the one on the right to make a single tif raster.

Hillshade has been applied using an ArcGIS Pro rasterfunction.

There are some strange effects going on around the edges of pixels in the left hand data. 

Brooks_SummitGeo_0-1625646150088.png

For clarity, if I change the resampling interpolation to bilinear, it smooths the pixels within each 50m group (100 pixel squares), and the strange effect goes away, but the 50m group of pixels is visible in the hillshade (indicating that the effectiveness of bilinear interpolation isn't quite perfect at smoothing).

Brooks_SummitGeo_1-1625647327583.png

If i change the resampling technique to cubic interpolation, the smoothing appears to be more continuous, but you see some strange gridding artefacts in the hillshade again, different to the nearest neighbour example.

Brooks_SummitGeo_2-1625647582499.png

Question to the raster pros out there, any thoughts?

PS: i've dropped these datasets into QGIS and Global mapper too, and the same hillshade effects are there too.


David
..Maps with no limits..
0 Kudos
1 Solution

Accepted Solutions
JayantaPoddar
MVP Esteemed Contributor

It's never a good idea to resample your data to a higher resolution, by 100 times. So, what you are basically seeing, are the hundred 5 meter pixels coming out of each 50 meter pixel.

The best approach that I can think of is 

1. Convert the 50 meter Raster to points 

2. Interpolate the Points to create 5 meter Raster dataset. Topo to Raster (Input Feature Data would be a Point Elevation)

Define the Snap Raster parameter (original 5 meter Raster) in Tool Environment.



Think Location

View solution in original post

4 Replies
DanPatterson
MVP Esteemed Contributor

Did you see the help topic on this?

Cell size and resampling in analysis—ArcGIS Pro | Documentation

 


... sort of retired...
0 Kudos
JayantaPoddar
MVP Esteemed Contributor

It's never a good idea to resample your data to a higher resolution, by 100 times. So, what you are basically seeing, are the hundred 5 meter pixels coming out of each 50 meter pixel.

The best approach that I can think of is 

1. Convert the 50 meter Raster to points 

2. Interpolate the Points to create 5 meter Raster dataset. Topo to Raster (Input Feature Data would be a Point Elevation)

Define the Snap Raster parameter (original 5 meter Raster) in Tool Environment.



Think Location
David_Brooks
MVP Regular Contributor

That's a great way to resolve the issue. I know it's sub-optimal to resample a raster in this way, but i need to have a composite dataset to work with at the smaller resolution. raster to points to raster seems spot on. Would you suggest IDW interpolation or something else?


David
..Maps with no limits..
0 Kudos
JayantaPoddar
MVP Esteemed Contributor

Topo to Raster should work fine here. But, you are free to test a few more interpolation techniques.



Think Location
0 Kudos