topic Re: Mosaic to New Raster input questions - using Mosaic Dataset in ArcGIS Pro Questions

Mosaic to New Raster input questions - using Mosaic Dataset

JasperRomero — Mon, 09 Feb 2026 19:55:35 GMT

Hello Community,

I recently created a Mosaic Dataset of ~50 historic aerial images. The mosaic dataset itself generally looks good and performs well as a hosted Tile layer on AGOL; however, to put the image on non-Esri GIS applications used by my institution I need to create a single raster file of the whole image. I am looking for clarification and guidance on options/best practices for creating this single raster file, mainly around ensuring resolution is as close as possible to the source images.

I have have experimented with using Mosaic to New Raster with the Mosaic Dataset as the input, and while this runs relatively quickly it appears to resample with a loss of resolution over the whole image. There is some variation in resolution among the source rasters for the Mosaic Dataset, ranging from ~0.72 - ~2 meters; however, the mosaic looks fairly smooth across the image (presumably because the Overview uses a single resampled resolution - is this correct?).

I would be thankful for any suggestions on how to create a single raster file that looks as close as possible to the Mosaic Dataset - and I am particularly curious about the following details:

1) Is it correct that the workflow as described above (Mosaic Dataset as input for Mosaic to New Raster) is generating the output raster from the Overview, rather than the individual source rasters? If so - is this contributing to the image degradation?

2) Is there any way to Mosaic to New Raster that mosaics using the Seamlines from the Mosaic Dataset? If the answer to 1 is yes, then I suppose that checks this box - but would there be other ways? I'd really like to preserve those seamlines in the single file output if possible, they significantly improve the quality of the mosaic.

Thanks in advance for any help or suggestions.

Re: Mosaic to New Raster input questions - using Mosaic Dataset

Robert_LeClair — Mon, 09 Feb 2026 20:56:50 GMT

In reading the Mosaic to New Raster (Data Managment) GP tool help, it does mention the inputs must have the same number of bands and same bit depth. In addition, it mentions if you don't set the bit depth, the 8-bit default will be used. The cell size paramter is likely the culprit of resolution variation as you did mention the cellsize varies from 0.72m to 2m. If you to go the ENVIRONMENTS of the GP tool, there is a resampling method you can change. It defaults to Nearest but there are other methods, bilinear and cubic that also may improve you output.

With regards to your questions:
1. Not to my knowledge, it uses the source rasters and not the overviews.
2. There is a Export Mosaic Dataset Geometry (Data Management)—ArcGIS Pro | Documentation that preserves the footprint, boundary, seamlines and spatial resolutions of a mosaic dataset. Maybe worth exploring.

Re: Mosaic to New Raster input questions - using Mosaic Dataset

JasperRomero — Mon, 09 Feb 2026 21:12:47 GMT

Thank you for responding @Robert_LeClair - All the source files for the Mosaic Dataset are 8-bit unsigned with a single band, so I have been largely ignoring those parameters so far but I will read up a little to see if they warrant a closer look.

Re: Cell Size - is it fair to say that the best bet is to set this param to the smallest (or... largest?) value of the input files and just accept that there will be some resampling (i.e. for the files with a different res?)? One of the people who requested the imagery will probably want to know why the resolution looks different between the AGOL hosted Tile layer and the file served out to other applications, so I'm trying to understand this well enough to explain why it (maybe) isn't possible for them to be the same.

I'll look into that GP tool as well - thank you!

Re: Mosaic to New Raster input questions - using Mosaic Dataset

Robert_LeClair — Mon, 09 Feb 2026 21:33:02 GMT

RE: the cell size. I'm not sure. 0.72 m to 2 m is significant range I suppose. You could experiment with a cell size in the middle, like 1.36m, and with the resampling it could work? Or try 1m and again see what the result is. Caveat - I'm not an imagery SME - just troubleshooting at the moment.