Best practices on the use of Raster and Mosaic Datasets, and file geodatabases

[brucesimonson]

Greetings,

I am looking for best practice advice on how to manage multiple tiled file sets of aerial imagery, and elevation DEMs, in the different formatting options suggested by ESRI. 

In particular, I'm interested in:

a)  if, when, and how (in particular) to use file geodatabases
b)  raster vs mosaic datasets

I have read this, and that, which are fine discussions in the 10.0 documentation on mosaic and raster datasets, and raster catalogs.  However, I'm still not clear on some ideas.

Here's a representative sample of my situation:

imagery:

3 bands - false color infrared - 9in pixels - 387 TIFs (as tiles), from 2001 - State Plane
3 bands - RGB - 6in pixels - 312 TIFs (as tiles), from 2006 - State Plane
8 bands - RGB+5bands - 2m pixels - 110 TIFs (TIFs overlap), from 2010 to 2012 - UTM
4 bands - RGB+IR - 6in pixels - 523 tiles of TIFs, from 2013 - State Plane

DEMS:

bare earth - 2 meter - 280 raster files (as tiles), from 2002 - State Plane
first return - 2 meter - 280 raster files (as tiles), from 2002 - State Plane
bare earth - 1 meter - 82 raster files (as tiles), from 2012 - UTM
bare earth - 3 feet - 472 raster files (as tiles), from 2013 - State Plane
hydroflattened bare earth - 3 foot - 472 raster files (as tiles), from 2013 - State Plane

I would like to optimize these (and other similar rasters) for use in ArcGIS desktop (ArcMAP) and for eventual deployment in ArcGIS Server.

Specific questions:

1)  I always wondered - since these are static, should I use file geodatabases?

2)  Should I create a separate file geodatabase for each product, or (somehow) put them all in a single, or at most a few, file geodatabases?

These products overlap much of the same area, but each product has its own specific coverage area.  Each product has its own tiling system as well.

Raster Datasets vs Mosaic Datasets

A couple of strategies come to mind:

Option 1:

  Create different raster datasets for each product, either:

       i) in the same file geodatabase, or
       ii) in separate file geodatabases (the above question number (2)).

Option 2:

   Same as option (1), but then create a single mosiac dataset out of these raster datasets.

Option 3:

  Create a single mosaic dataset for everything, in a single file geodatabase, but using the original tifs.

Option 4:

  Create different mosaic datasets, based on product (e.g., same number of bands in the imagery, same DEM type, etc)?

---

More questions:

3)  Does Option 1 save space?  Seems like the raster datasets I've made of these TIF tile datasets are considerably smaller than the original set of TIFs, byte-wise.  I think once the raster datasets are created, I can use them, and can archive the original data in off-line storage. 

4)  Does Option 3 (mosaic only) require the original TIFs and DEMs to be stored where the mosaic can find them?  I guess what I mean by this, is, does the mosaic contain pointers to the original files, or do the mosaics make their own copy?  If there are pointers involved, then the original data must be available online, I reckon.

5)  Does one of these approaches work better for eventually caching these products, for deployment on ArcGIS server?

6)  There are two different projections in play in these data -- should I project everything into one, before I start, or will ESRI's on the fly projection capabilities take care of this for me?

---

I know this is a lot; it's overwhelming, just typing it up.  What should I do?  What would you do?  Option 1, 2, 3, or 4?  Or something else?  Any obvious gotchas with any of these approaches?

Cheers, and thanks for your thoughts,
-Bruce

[TerryOD]

Bruce, would be interested to know what route you took, we have similar questions!!