Smoothing seam between adjacent DEMs

[FrancesBiles]

Hello,

I have three exactly adjacent DEMs I am trying to mosaick together.  There are no overlapping or NoData areas between the DEMs. The DEMs are in the same coordinate system, have the same vertical units, same cell size (50 meters), and are snapped to the same grid alignment. The DEMs are all terrain models (not surface models) and are stored as ESRI GRIDs. One of the DEMs is from the U.S. (NGVD29), and two are from Canada (CGVD28). The slight difference in vertical datums and source agencies contributes to variability in the elevation values between the data sets and creates a distinct "seam" where the DEMs are joined. 

What is the best way to smooth the seam where the DEMs are mosaicked/merged? I have created some polygon feature classes of varying widths that overlap the seam area with the idea of using one as a mask for applying some kind of smoothing function. Am I on the right track?  Specifically:

1) Is there an efficient way to output the whole, merged DEM with the seam area smoothed? and
2) What are the best functions and options to be using for smoothing continuous elevation data (e.g., focal mean? neighborhood shape and size?).

I have found several answers addressing smoothing overlapping areas between DEMs/rasters and how to fill missing data areas, but have not found anything addressing this issue.  I would be grateful for any advice or pointers to other links on this topic.

Running ArcGIS 10.0 with an ArcInfo license.

[CarrieDavis · Accepted solution]

Hi,

Since you have an ArcInfo license, you may want to look into creating a mosaic dataset.  With a mosaic dataset there are several mosaic and stretch methods you may experiment without needing to create one large DTM for the whole area.  You may also use the mosaic dataset in many of the geoprocessing tools as a raster input and/or copy it to a single raster format using the Copy Raster tool.

Workflow to create a mosaic dataset
http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#/The_workflow_and_tools_to_create_mosai...

Below is also a knowledge based article on how to remove and replace nodata values with surrounding values. You could use also just use the focal statistics tool on its own to smooth the final raster.

HowTo:  Remove and replace no data values within a raster using statistical information from the surrounding data values.
http://support.esri.com/en/knowledgebase/techarticles/detail/21261

Focal Statistics (online documentation)
http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//009z000000qs000000.htm

Best wishes.

[FrancesBiles]

Thank you for the reply Carrie.

I do not want to use a mosaic dataset because I need to create a layer to use in regional models and analyses and to share with others that may or may not be using ESRI tools. Also, as mentioned above, filling "no data" areas was not an issue.  However, I did end up using focal statistics to smooth out the seam (so marked this question as answered) using the following syntax in Map Algebra:

Con(~IsNull("SeamMask"), FocalStatistics("DEM", NbrRectangle(5, 5, "CELL"), "MEAN"), "DEM")

Where SeamMask is a 200 meter wide raster that overlaps the seam where the DEMs adjoin.

The results aren't fantastic, but it does help ease the transition between the DEMs a little.