search

Esri / ops-server-config

Operations Server Configuration scripts
Apache License 2.0
34 stars 13 forks source link

Elevation services have raster functions that do not function correctly in Geographic WGS84 #320

Closed conklinbd closed 9 years ago

conklinbd commented 9 years ago

The current set of elevation services have a number of raster functions (Slope, Hillshade etc...) That do not function correctly if the service is in WGS84 Geographic. @nfeuerstein @NatalieCampos @mfunk, let's have the discussion here and not in email where the decisions made are getting lost.

conklinbd commented 9 years ago

From @nfeuerstein Perhaps I misunderstood what Vinay was saying and our follow-up conversation after the meeting, so please tell me if I’m mistaken.

I thought we were offered the following solutions:

  1. Start with WGS84 and add a reproject in the function chain before the slope function.
  2. Continue to use Web Mercator and understand that a CR is being logged.
conklinbd commented 9 years ago

From @NatalieCampos

The work around with the Reproject function has to be tested for accuracy. Mike was not a supporter of this method or any of the work arounds presented. And then comes the issue of is using the Reproject function, is that what we want to represent as “best practice”?

So I can test the reproject function and see if it give me plausible results or we can revert back to web Mercator. I vote for reverting back while we do the testing.

nfeuerstein commented 9 years ago

If the workaround (reproject in the raster chain before the slope/aspect/hillshade function) provides accurate enough results - at least as accurate as web mercator - then this seems a viable option until a CR has been fulfilled. Any grievances with this?

@NatalieCampos please provide results from your tests and then we can make a decision.

conklinbd commented 9 years ago

This is good with me. I think it is a viable option for our customers.

NatalieCampos commented 9 years ago

I used a Landsat 8 scene of Barrow Alaska to perform the test to see if there is an obvious visual shift in the data when the reproject function is applied to a wgs84 dataset. See the pics below.
image

nfeuerstein commented 9 years ago

So, yes? The visual shift is too great and reprojected WGS84 provides a less accurate result then web mercator?

conklinbd commented 9 years ago

It looks good to me. The point shows up in the correct location in both views.

nfeuerstein commented 9 years ago

One dot is blue and one dot is purple. @NatalieCampos - are the blue and purple dot the same dot?

NatalieCampos commented 9 years ago

Natalie

These are the same dot. These are the things to consider:

  1.   The image on the Left is in WM.  The image on the right is in WGS84.
  2.  The dot is in WGS 84 so is native in the WGS image but the dot is being reprojected on the fly for the WM image.  It is an issue. I would need to check for a data shift in a different software not ours as we always reproject.  But I think it is close enough.
  3.  The image itself is of Barrow Alaska. Way up north.  Which is why the WGS84 image looks horribly distorted.  Given  how north Barrow is and the two linked maps in two different projections.  The visual extent is fairly consistent and not vastly off as seen in the NGA’s document.  This would mean that we are visualizing the correct rows/columns of the image or there is a project on the fly happening when we link.

As of now, I am going to assume that using the reproject raster function doesn’t produce a shift in the data and can be used with the Aspect/Slope/Hillshade function as a work around until the issue is resolved in core.

NCampos

From: Natalie Feuerstein [mailto:notifications@github.com] Sent: Tuesday, September 09, 2014 3:15 PM To: ArcGIS/ops-server-config Cc: Natalie Campos Subject: Re: [ops-server-config] Elevation services have raster functions that do not function correctly in Geographic WGS84 (#320)

One dot is blue and one dot is purple. @NatalieCamposhttps://github.com/NatalieCampos - are the blue and purple dot the same dot?

— Reply to this email directly or view it on GitHubhttps://github.com/ArcGIS/ops-server-config/issues/320#issuecomment-55044264.

NatalieCampos commented 9 years ago

After speaking with Nathan, the linking does reproject so that is why they are lining up. Shall we move to a different software or shall we proceed with the work around?

nfeuerstein commented 9 years ago

To make sure I understand: The dot is WGS84. When displayed on the web mercator image it is reprojecting-on-the-fly to web mercator?

NatalieCampos commented 9 years ago

Yes.

nfeuerstein commented 9 years ago

So thats why they line up? Because our software is awesome. Do you have access to another software suite that does not reproject on the fly, so that we can visualize if there is in fact a shift?

conklinbd commented 9 years ago

If it works correctly in our software as an image service (in both web a desktop) then we should consider this fix good.

NatalieCampos commented 9 years ago

raster functions were created that include the reproject to web mercator and attached to the DTM service