Working with Delaware 2007 Aerials in ArcGIS

Well, the 2007 orthophotography for Delaware have been released and are making their way around the state. I just started working with them and thought I would note a few things while working in ArcGIS.

First, a few tech specs:

• 2010 GeoTiff files, including world (.tfw) files
• tiling structure of the 2002 orthos (see http://maps.rdms.udel.edu/orthoindex/)
• horizontal resolution is 0.25 meters
• each tile is 6800×6800 pixels; 1.7 km x 1.7 km
• 4 bands: Red, Green, Blue, Near-Infrared
• file size ~ 180 MB; pyramids (rrd files)~ 15 MB

While viewing these in ArcGIS Desktop, you probably have the software configured to display the first three bands by default. This works out well in these files for this displays true color imagery. In the ArcMap layer properties:


True Color
Red = Band 1 (Red band)
Green = Band 2 (Green Band)
Blue = Band 3 (Blue band)


However, to view a false color composite of the NIR information, set the band combination as follows:


False Color IR
Red = Band 4 (NIR band)
Green = Band 1 (Red band)
Blue = Band 2 (Green band)


The 2007 imagery does look as intense as do the 2002 or even the 1992 false color. This could be due to:

• differences in spectral windows of the sensor being used
• image preprocessing; 2007 data was captured 100% digital while 1992 and 2002 were captured using IR film and scanned
• 2007 imagery was taken leaf-off, as it should, about three weeks earlier (I think) than the 2002 imagery
• the software is not stretching the brightness data like you are accustomed to

As delivered, statistics are NOT calculated for each image. Many software packages, like ArcGIS, utilize pixel statistics to visually stretch the image in various ways. Being GeoTifs, stats can lie inside the header of each TIFF file, or externally in .aux files. ArcGIS will create stats in an .aux file, which it uses to store other types of information about the image. If stats do exist in the header, I do not believe ArcGIS Desktop can take advantage of them. (I need to test this more.) You will need to use ArcToolbox (Data Management, Raster, Build Statistics) to create the statistics. Once created, open the layer properties and use a 2 standard deviation stretch for starters.

Also as delivered, pyramid files are not created. These dramatically help performance when displaying the imagery at various scales. Similarly to stats, in GeoTiffs, pyramids can live inside the header or as external .rrd files. ArcGIS and other software (like IMAGINE) use .rrd files. Many open source packages can use the pyramids (also called overviews) in the file header (I do not know if ArcGIS can.) ArcGIS will probably build them for you automatically when viewing in ArcCatalog or ArcMap. If not, you will need to use ArcToolbox (Data Management, Raster, Build Pyramids) to create the pyramids.

Lastly, the coordinate system is defined in the GeoTiff header, but it is incorrect. It is listed as NAD83 State Plane Meters Delaware FIPS 0700. However, the imagery was georectified using ground control points surveyed to align with the HARN network. These were the same points used for the 2002 imagery. Therefore, both the 2002 and 2007 imagery, as well as any data (vector or raster,) that was georeferenced from these. ArcGIS can read the CS info from the GeoTIf header. It may even write some info to the .aux file (I’m not sure.) You can define/modify the coordinate system in ArcCatalog or ArcToolbox (Data Management, Projections and Transformations, Raster.)

To help with all of the above, you can runa third party script to modify all three parameters at once: stats, pyramids, coordinate system. I used the following script with various degrees of success. This script will go through an entire directory of rasters and mofidy any or all three of these parameters.

Raster Spatial Reference Tools
http://arcscripts.esri.com/details.asp?dbid=12820

I have also used the open source GDAL library and utilities for working with the imagery, but that’s for another post….