Rasters - correcting overlaps

[BruceCampbell]

Hi -- I'm working with multiple stacked rasters that represent the tops of aquifers and confining units. The rasters were constructed from a combination of picks on boreholes and a land surface DEM. The problem is they overlap in places and will not work in the groundwater flow model that I'm constructing. Is there a way to "correct" the overlaps in the rasters in ARC? I suspect the overlaps occur during the interpolation process of the XYZ data.

Any ideas would be appreciated!

[IanMurray]

Hi Bruce,

If you have multiple overlapping DEMs, it might be good to Mosaic them together into a single image.   There are many tools for doing this, but you probably want to keep your original data the same, I'd suggest Mosaic to New Raster.  That way none of your originals get changed, and you have a new single DEM file to work with.   If there are differing values for the same location, you can control how you want them to combine in the mosaic_method parameter.

ArcGIS Help (10.2, 10.2.1, and 10.2.2)

[DarrenWiens2]

As you allude to, the edges of interpolated rasters are the most distorted because they are only influenced by data on the inside.

You can try to minimize the overlap distortions using a mosaic operator (like blend) that will combine the overlapping distortions. Or if the distortion area is uniform around each raster, Shrink the rasters by a number of cells to reduce the area of overlap, and recombine using the more reliable parts of the rasters (i.e. more of the "insides").

Of course, your best option would be to obtain the original data and reprocess it into one continuous coverage, but I'm guessing that's not an option.

[CodyBenkelman]

Bruce

can you expand on what you are seeking when you say "correct" the overlaps?  Do you need to change which is on top, or are you having problems where edges mismatch the underlying data?

The Mosaic Dataset is specifically designed to combine rasters, even if they are at different resolutions and projections.  The Mosaic Dataset (MD) is a virtual raster - a database table that points at your source files but does not copy the pixels - and it may or may not work in your groundwater flow model.  But it has a number of options for deciding which pixels appear on top in areas of overlap - e.g. you can dynamically reorder the rasters by any available attribute (date, resolution, accuracy, etc.) and that may be of interest if you want to run your model with different rasters as "first priority".  Or if you want to blend overlapping pixels etc. 

You did not mention NoData regions, but I'm wondering if that is a related issue you'd need to manage...?   At version 10.3, ArcGIS added a function called "Elevation Void Fill" that provides an option to dynamically fill NoData areas if that is an interest.

If your model cannot take a MD as input, I'd suggest you manage the various rasters in an MD and then once you have the overlaps properly addressed you can output a raster at any desired resolution & projection.

If you could describe your data & requirements a bit further, and let us know your version of ArcGIS, we may be able to advise with more specifics.

Cody B

[BruceCampbell]

Hi Cody -- Thanks for the quick response and helpful information. I'll try to describe what we're doing and the rasters in a little more detail. The raster surfaces represent the tops of 8 hydrogeological units. The study area is in the Upper Coastal Plain of South Carolina. The units dip toward the Atlantic Coast and all are exposed (to some degree) at land surface. The upper units have the largest areas of exposure with the deeper units having less exposure. We are using the USGS DEM for land surface altitudes. The subsurface "picks" on the tops of the units are mostly done from boreholes but some are from outcrops in geological "windows" where streams have eroded some or all of the Coastal Plain sediments away and exposed bedrock or the deeper units.

We created the rasters from shapefiles of the areal extents of the hydrogeological units and the altitudes of the borehole picks and outcrops. However, we think the interpolation process must be giving us the "crosses" - areas where what should be a lower hydrogeologic unit is above an upper unit. We put the raster in ARCScene, separated them and can clearly see the problem areas.

If we can correct these areas, we'll be able to use the individual rasters directly in the GUI for the model setup. We won't be able to use the mosaic directly in GUI. We're using ARC 10.2.1 but I plan to get the 10.3 upgrade this week.

Thanks again for the help.

[NICOLAPEDDIS]

Bruce, check this usefull pdf

http://pubs.usgs.gov/of/2014/1167/pdf/ofr2014-1167_carrell-tools-and-techniques.pdf

other useful tutorials are present over the web.

I'm try do do your same job in a new project, and now i'm collecting the boreholes and the acquifers to my watershed , with a 10 meters DEM

[BruceCampbell]

Hi Nicola -- Thanks, the tool and techniques PDF file looks very useful.

It's a challenging process to put all of this together but looks very

useful when completed.

[NICOLAPEDDIS]

Yes Bruce, it's a really hard work to accomplish, due the limitations of arcscene,

An now i'm spending more time to read and understand the tecniques.

king luck for your excellent work..

[CodyBenkelman]

Nicola

thanks for that interesting PDF.  It appears the hyperlinks are no longer valid; have you already contacted Jennifer Carrell (the author) to ask about updates?  If not, I'll ask and post the results.

Thanks

Cody B.

[NICOLAPEDDIS]

Cody, not yet, and i know that the hyperlinks are not valid, You need to search the tool through the Illinois State Geological Service/Survey. That's what i did..