03-20-2023 11:38 PM
New Contributor

Does anyone have an update for this BLOG on the Esri Australia site?

The reference details that the NTv2 grid for ArcGIS Pro is useful for standard GDA2020 web maps but should not be used when mixing datums and Web Mercator Basemaps in a publishing workflow.

The blog is two years old and we are wondering if there is any change to the available conformal transformations and Web Mercator / GDA2020





0 Kudos
2 Replies
MVP Esteemed Contributor

You would be advised to contact esri tech support, but... this blog is more recent, just slightly

New GDA2020 transformations for WebGIS | Esri Australia Technical Blog (

... sort of retired...
0 Kudos
MVP Regular Contributor

Hi, I'm the original author of that post.

Publishing to ArcGIS Online will do one of two things:

  • Project (and transform) any data in the map to align with the map spatial reference, using the transformations set at the map properties.
  • Publish the service with any supported transformations defined at the map level i.e. GDA2020 data + map. These transformations stick in the service and will be invoked when added to a web map. I.e. WGS84<>GDA2020 transformation. If none is defined, ArcGIS Online will use the default (currently WGS84 <> GDA2020 2, a null transformation that aligns data in WGS84@GDA2020).

ArcGIS Online doesn't support grid-based (NTv2) transformations for Hosted Feature Layers. Only the default. ArcGIS Pro, ArcGIS Server and Field Maps can support the grids if they are installed. They are large, and a lot of countries have them, hence the limitation. If you publish a hosted feature layer with a NTv2 grid-based method, it will be dropped from the service definition.

So if you want to use a grid-based transformation, you should do it BEFORE or DURING publishing to a Hosted Feature Layer. i.e.

  • Project the data to a new dataset using the grid-based transformation method, then publish
  • Set the map reference to the new spatial reference (e.g. GDA2020), add a grid-based transformation method to the map, and then on publish the GDA94 data will use the transformation method of the map to align the data to the map spatial reference (i.e. projecting + transforming on publish).

As Dan mentioned, there are new transformations now. Originally they didn't release any direct transformations between GDA2020 and WGS84. Now there is, and they are poorly named in my opinion. Key to note:

  • GDA to WGS84 is 'null'. They just dump the data. Originally this dump was in line with GDA94. Now we can dump the data in line with GDA2020. So all 'WGS84' spatial references are actually 'WGS84@GDA94' or 'WGS84@2020' - a representation of either 94 or 2020 in WGS84.
  • All the transformation methods will do nothing, or apply a transformation between 94<>2020 prior to WGS84. So you need to pick if you want to align in WGS84@2020 or WGS84@94, then choose a transformation path to achieve that. e.g. WGS84@GDA2020:
    • You can choose 'WGS84<>GDA2020 2' for GDA2020 data, as this is a null transformation
    • 'GDA94 to WGS 84 (2)' will transformation to GDA2020 on the way, aligning GDA94 data to WGS84@2020.
    • Some transformation methods take the default method, others use the grids, so if you choose a method with a grid - keep in mind the earlier comments.
  • If you have data aligning to WGS84@2020 and WGS84@94 in a web map, they will NOT align. If these layers are actually in WGS84 (vs 94/2020 + transformations), you will never be able to get them to align in a web map as you can't update the transformations at the service level. they are in the 'correct' WGS84 spatial reference.

Personally if I am publishing to ArcGIS Online I would:

  • Align to 'WGS84@GDA2020' unless there is a need not to i.e. existing content
  • Project before or during publishing, so that the data is not projected on the fly in the Web Map.
  • Include a metadata comment in the service to note that the data source spatial reference + the method used to transform it (otherwise it's just WGS84 and confusing)
  • Decide if a grid-based method should be used and ensure that ALL data is transformed with the same method + I am aware of where these methods cannot be used. Which grid-based method (i.e. distortion or no) depends on your state/region.

Any other questions feel free to post, might write an updated blog this weekend.