Best Practices for Imagery Basemap

[PaulVanDyke]

I'm looking for suggestions for the best way to create an imagery basemap.  I have a bunch of images that I'd like to combine into one.  They are from 2004 thru 2010 (currently, with some 2013 expected shortly).  They vary in resoluiton from 1 meter to .5 foot.  I'd like to create two seperate basemaps.  One should be focused on the best resolution and the other on the most recent imagery.

Unfortunately my imagery look something like this:
2004 - 1 meter
2008 - .5 foot, 1 foot, 2 foot
2009 - .4 meter, .5 meter
2010 - 1 foot, .5 meter

Not all the images are overlapping.  Many of the higher resolution images have a smaller footprint inside the lower resolution images.

You can see a start of this at http://gis.kodiakak.us/arcgis/rest/services/Imagery_Basemap/MapServer?f=jsapi

I created that map service using the raster catalog tool and importing non-overlapping rasters.  Now I need to add overlapping rasters to the mix.

Thanks,

Paul VanDyke
Kodiak, Alaska

[CodyBenkelman]

Paul
I'd encourage you to use the Mosaic Dataset - it's dramatically improved over the raster catalog. 

Create a separate "source" mosaic dataset for each year, and when you add rasters, have Update Cell Size Ranges on (default) to calculate visibility based on pixel size.  Add a field into the Attribute Table for "year".  You'll have to populate that field manually.  If your different years are in different projections, create each source mosaic dataset in the original data projection.

Then create a new mosaic in the projection for the base map (web mercator aux. sphere?) and input the single-year mosaics, being sure to use RasterType = TABLE, and also be sure Update Cell Size Ranges is OFF (so the visibility vs. resolution is maintained from the original source mosaics).  We refer to this as a "derived" mosaic dataset, but note you create it just like any other. (It's not fundamentally different like a Referenced mosaic dataset; you should learn about those, too, if you haven't used them)

I see you're on 10.2, so from the derived Mosaic Dataset, you can create image caches using ArcGIS Desktop and those cache datasets will be your base maps.

When viewing or creating a base map cache, set the Mosaic Method to "By Attribute", and select "LowPS" --> this will display the highest resolution image in any area.  Alternatively, with Mosaic Method to "By Attribute", and field "year" (DESELECT "order ascending" since when mosaicking by attribute, smaller numbers have higher priority - otherwise the OLDEST images will appear on top).

This should work, *unless* you're going to cache your base maps at 0.5 feet, in which case the software will automatically put the highest resolution images on top (even if not the most recent) when you zoom to full resolution in the "newest on top" version.  Let me know if you need more guidance on this issue.

Note Mosaic Method is a property of the Mosaic Dataset; you can set the default by right clicking the MD in Catalog, or after loaded into the map, change it by right clicking the Image layer of the MD in the table of contents (see "Mosaic" tab).

I think I've covered everything, but test it and if something doesn't work, let me know.

[PaulVanDyke]

Thank you for the input.  I have started down the road with the following steps:

1) Create file .gdb for housing rasters, feature classes (for clipping) and each primary mosaic dataset.
2) Populate each file .gdb with the relevant raster dataset
3) Add raster to the mosaic dataset, build footprints, boundaries and overviews
3) Map out clipping polygon as feature class in the file .gdb
4) Apply raster clip function to the mosaic dataset
5) Add each primary mosaic dataset to a derivative mosaic dataset
6) Modify the MinPS and MaxPS values to show all imagery at all levels
7) Set defaults to order imagery by attribute, ZOrder
😎 Modify the ZOrder atrribute for each entry in the derivative mosaic dataset to correctly order the imagery
9) Export the whole thing as a tile package to be published on our ArcGIS Server

Unfortunately, I think I'll be re-doing this process from scratch AGAIN...

I'm exporting the tile package in the ArcGIS Server/Bing/Google WGS84 Web Mercator (Aux Sphere) format and I'm seeing the initial results and they look a little jagged and blocky.

The original source imagery is NAD83 Alaska State Plane Zone 5 (feet) or NAD83 UTM 5N (meters) and it was added to the mosaic datasets in the original formats.  I've never been impressed when transforming imagery from this format to WGS84 WMAS on the fly.  I've learned by experience that the best thing to do it to transform the data into WGS84 WMAS at the beginning and then create the pyramids on the data in it's new projection.  I have been using "bilinear" for all my imagery pyramids.

I'm doing all this work on my GIS workstation (a Dell 3600 workstation-class machine) w/ 32GB ram and it looks like the tile package export will take a few days for an extent about 60k+ sq. km.

Any input is greatly appreciated,

Paul VanDyke