Process_Step:
Process_Description: The 7.5 minute image files were copied from the distribution
CDs to a disk drive. Images were selected for registration quality
checking using a stratified random sampling scheme in which two
non-adjacent maps were selected from each 1 degree block that was
entirely within New York and one map was selected from each block that
was partially within New York. This approximately 3% sample was
evaluated by checking the neatline corners of the scanned map against
a vector coverage of the quadrangle boundaries. Only three of the 47
sampled maps had neatline corners within one pixel of the vector
boundaries. The remainder were mis-registered by between one and five
pixels, equivalent to between 2.4 and 12 meters, in one or both
directions. We decided, therefore, to re-register the images.
Information on the evaluation process can be found in the report
Evaluating DRG Image Registration Accuracy. Our ultimate goal was a
set of images in UTM zone 18 in NAD83. A set of tests demonstrated
negligible differences between the computationally time consuming
process of projecting the images and projecting the coordinates of the
upper left "placement" point in the ESRI world file. Theoretical
considerations indicate little difference between projecting the image
and using the ArcInfo GRIDWARP command to re-register the images
because the NADCON grid spacing is 15 minutes. We decided to use
GRIDWARP to perform datum conversion and re-registration at the same
time. GRIDWARP was set up by creating links from the four corners of
the image's map neatlines to a vector coverage of quadrangle
boundaries. The vector coverage was derived from a coverage that had
been generated in geodetic coordinates and projected to UTM Zone 18 in
NAD27. For DRG re-registration this coverage was projected to NAD83
and the proper UTM zone for the quadrangle. Each image was converted
to a grid, subjected to a first order (linear) GRIDWARP, and converted
back to an image. The registration quality at the four corners was
checked by a different person. The registration process was repeated
if necessary until a good match was obtained between the raster and
vector quadrangle corners. Eight quadrangles (o41076h5, o44074b8,
o40073h2, k44073c7, k44073c5, o42073d7, k42073e3, o42073h7, o42079a1,
and o42079a2) could not be successfully registered. Additional links
were added for these files and they were registered with a second
order GRIDWARP. Map collars were clipped from the images by extracting
the zone and bounding geodetic coordinates for each quadrangle from
the .fgd files that accompanied the images. This information was used
to create a coverage that was projected from NAD27 to NAD83 in the
proper UTM zone and used to clip a grid made from the image. The grids
were then merged. Merging was automated for up to 48 images at a time.
These grids were further merged. Grids with all quadrangles in UTM
zones 17 and 19 were created and projected to zone 18. We attempted to
create a single grid for the entire state but were unable to do so
because the result would have exceeded the 2.1 GB size limit for a
single grid. We therefore somewhat haphazardly merged the grids into
larger grids. We then divided the state into six longitudinal zones.
An outline coverage for each zone was created. All sides of these
grids were oriented along the coordinate axes so that no areas of
NODATA would be created when they were used to clip grids. Each
outline coverage was then used to clip each of the larger merged
grids. Finally, these clipped component grids were merged to grids
that exactly filled a longitudinal zone. We created a coverage of
polygons that were to be used to define the final image tiles. These
polygons were squares of 5000 m oriented along the coordinate grid.
The tiles were designed to coincide with the blocks used by the New
York State Breeding Bird Atlas. The minimum X coordinate of the
Breeding Bird Atlas coverage was projected to NAD83. This value was
decreased by 20,000 and used as the starting coordinate in an ArcInfo
GENERATE with FISHNET to produce square polygons 5000 m on a side.
This coverage was then clipped by a polygon created by buffering a New
York State outline coverage by 24,200 m. Labels were copied from the
Breeding Bird Atlas coverage. Additional labels were added to the
polygons that did not occur in the Breeding Bird Atlas coverage. These
polygons were given a unique value for BLOCK, the item in the Breeding
Bird Atlas coverage that identified each 5000 m square. The 5000 m
square polygons were used to clip 5000 m square tiles from the
longitudinal zone grids. Each 5000 m square polygon was extracted into
a separate coverage. The minimum and maximum coordinates of this new
coverage were determined. The maximum coordinates were incremented by
2.5, the approximate size of pixel, to deal with improper rounding in
the ArcInfo GRIDCLIP command. These coordinates were then used as
arguments to the Grid GRIDCLIP command with the BOX option to clip a
5000 m square tile out the longitudinal zone grid. This grid was then
converted to an image with IMAGEGRID. Image file names were obtained
by preceding the BLOCK item from the coverage PAT with "o" and
converting the letter in BLOCK from upper case to lower case. An image
catalog was created to manage the images. The AML FILELIST function
was used to create a file that listed all the images. The name of this
file was used as the argument in the ADDIMAGE command in order to
populate the image catalog.
Process_Date: 19991231