5.B.ii. Global Data Set Assembly

GLOBE's final assembly was created in GRASS' r.patch module. This module combines data of various grid spacings and coverage areas into a resultant data set of defined coverage area and grid spacing. GRASS lets one work within a defined mask. It first inserts the highest-priority data set into the output grid, then replaces all remaining areas with a “0” value with values from the next-highest priority data set, ad infinitum. Working masks can be changed at any point in the sequence; this requires restarting r.patch.

The first task was to determine the priority of patching. The second task was to determine the mask(s) to be applied at each stage of patching. As GRASS version 4.1x uses zero as its "no-data" mask, and cannot distinguish between numerical zeros and areas of "no data," one must design masks to compensate for this characteristic of the software. Subsequent versions of GRASS, currently under development, are not subject to this characteristic. However, GLOBE Version 1.0 was managed under GRASS 4.1.4, which does have this characteristic.

The following patching sequence was used:

a. Insert Japan-GSI. (No mask applied during the patch.)

b. Insert Italy-SGN. (Japan-GSI coverage mask applied, but this is trivial.)

c. Insert Australia-AUSLIG/NGDC to B.A.D. GLOBE. Insert Australia-DCW/USGS/GTOPO30 to G.O.O.D. GLOBE. (Mask accounting for previous steps applied, but this is trivial.)

d. Insert Antarctica-SCAR/USGS/GTOPO30 as repaired by NGDC (now NCEI). (Mask accounting for previous steps applied, but this is trivial.)

e. Insert for conterminous U.S. and vicinity:

i. NIMA spot data. (Mask accounting for previous steps applied, but this is trivial.)

ii. DMA/NGDC data. (Mask accounting for previous steps applied, now nontrivial. This mask prevents overwriting of inland areas at sea level, but allows non sea-level values in DMA/NGDC data to overwrite sea-level values in NIMA spot data in coastal areas.)

iii. DMA/USGS/GTOPO30 data (Mask accounting for previous steps applied. This mask acts as the previous mask, allowing coastal zeros to be overwritten by coastal non-zeros.)

f. Insert NIMA spot data within 50^o of the Equator. (Mask accounting for previous steps applied.)

g. Insert NIMA/GTOPO30 data poleward of 50^o of the Equator. (Mask accounting for previous steps applied.)

h. Insert DEM for Greenland, including Zwally (and others)/NSIDC/JPL-DCW blend. (Mask accounting for previous steps applied.)

i. Insert DEMs from cartographic sources. (Mask accounting for previous steps applied. This allows DEMs from cartographic sources to line some coastal areas, such as limited areas in Greenland.)

Plate 1 is a histogram of elevation distributions in GLOBE. Prominent features in this histogram can be traced to features within individual sources of data. Histograms of these data are plotted in Plates 2 - 30 and discussed in Section 5.A. For example:

• Spiking is more prominent at elevations of 4000m and below, as few data from cartographic sources appear to cover areas with elevations higher than 4000m.
• The abrupt falloff in values above 4000m visible in the Antarctic Digital Database (Plate 30) is prominent in Plate 1.
• The bulge in elevation values around 5000m visible in DTED discrete coverage of Asia (Plate 3) is visible in Plate 1.
• The bulge in elevation values around 2500 - 3000m visible in ADD (Plate 30), and Greenland (Plates 20 and 21) is visible in Plate 1.

Inspection of the resultant DEM shows a relatively good fit between pieces, though there seems to be a modest disagreement between DCW-derived grids and DTED over vertical datum. In addition, the forcing of 0 elevations to 1 in categories used in GTOPO30 may cause difficulties for some users. However, discrepancies appear to be in this order of subtlety.