ETOPO2v2 2006 (version: ETOPO2v2c - Cell-centered registration)

Data set credit:

U.S. Department of Commerce
National Oceanic and Atmospheric Administration
National Environmental Satellite, Data, and Information Service
National Geophysical Data Center
http://www.ngdc.noaa.gov/mgg/fliers/06mgg01.html

ETOPO2v2c Data structure and provenance:

ETOPO2 grids are in meters.

Complete ETOPO2v2c grids are under folder
http://www.ngdc.noaa.gov/mgg/global/relief/ETOPO2/ETOPO2v2-2006/ETOPO2v2c/
as follows:

netCDF/
ETOPO2v2c_f4_netCDF.zip - 4-byte floating pt grid in COARDS-compliant netCDF Format

GRD98/
ETOPO2v2c_i2_LSB_GRD98.zip - 2-byte integer Most Significant Byte First (bigendian) grid in GRD98 Format
ETOPO2v2c_i2_MSB_GRD98.zip - 2-byte integer Least Significant Byte First (littleendian) grid in GRD98 Format

raw-binary/
ETOPO2v2c_i2_LSB.zip - 2-byte integer grid with Least Significant Byte First (littleendian)
ETOPO2v2c_i2_MSB.zip - 2-byte integer grid with Most Significant Byte First (bigendian)
ETOPO2v2c_f4_LSB.zip - 4-byte floating pt grid with Least Significant Byte First (littleendian)
ETOPO2v2c_f4_MSB.zip - 4-byte floating pt grid with Most Significant Byte First (bigendian)

HDF/
ETOPO2v2c_HDF.zip - HDF Format

First and foremost, ETOPO2v2c eliminates a 1-cell westward bias that was present 
in ETOPO2 as a result of alignment adjustments to match the choice of grid
registration used in sampling GLOBE data.

ETOPO2v2c is in a cell-centered registration (pixel registered); the cell
boundaries are lines of even minutes of latitude and longitude, and they are 
centered on intersections of lines of odd minutes of latitude and longitude. 
This is a change from the original ETOPO2, which was grid-registered, meaning 
that cells were centered on the integer multiples of 2 minutes (even minutes) of 
latitude and longitude and the cell boundaries were consequently defined by 
lines of odd minutes of latitude and longitude. The first row of 10,800 (360 x 
30) cells in ETOPO2 redundantly repeated the same data value, centered on the 
north pole. A data row, centered on the South Pole, was not included in the 
original ETOPO2.

Coverage of ETOPO2v2c is -90deg to +90deg in Latitude, and -180deg to +180deg in 
Longitude, whereas ETOPO2 covered +90deg to -89deg58' in Latitude and -180Wdeg 
to +189deg58' in Longitude. The new ETOPO2v2c, with cell centered-registration, 
eliminates the ETOPO2 North Pole redundancies but has the same number of cells: 
5,400 rows of data (180 x 30), each with 10,800 columns of data(360x30).

Program names in the following text, such as "xyz2grd", "surface", and 
"grdsample", refer to programs in the free set of Generic Mapping Tools (GMT), 
available at
     http://gmt.soest.hawaii.edu/

The GMT manual is available online at
     http://gmt.soest.hawaii.edu/gmt/gmt_man.html, in which usage and parameters 
are defined.

Smith-Sandwell (SandS) data were processed from a 1-minute grid-registered data 
file provided by Walter H. F. Smith (NOAA-NESDIS) and verified by Karen Marks 
(NOAA/NOS), available at
     ftp://falcon.grdl.noaa.gov/pub/walter/Gebco_SandS_blend.bi2
This is a binary integer grid of edited elevation data derived from the original 
interpretations of satellite altimetry available at
     http://topex.ucsd.edu/marine_topo/mar_topo.html

The downloaded file is converted to a GMT grid-registered grid by the GMT 
program xyz2grd, which processes the input file Gebco_SandS_blend.bi2 into the 
output file Gebco_SandS_blend_1m.grd:

xyz2grd Gebco_SandS_blend.bi2 -GGebco_SandS_blend_1m.grd=2 -R-180/179:59/-90/90 
-fig -I1m -ZTLh -V
where:
     the -G prefix identifies the output file name and the =2 suffix directs the 
program to create a 2-byte integer output file
     the -R prefix identifies the geographic region for the data by its W/E/S/N 
limits in degrees[:minutes[:seconds]]
     the -fig group identifies the (i)nput file as being in (g)eographic 
coordinates
     the -I1m group sets the (I)nterval between data points as 1 (m)inute
     the -ZTLh group specifies the input data as Z (elevation) only, arranged in 
rows (T)op-down and (L)eft-to-right in (h)alf-word (short) binary integers
     the -V requests (V)erbose output from the program directed to the Unix 
channel stderr.

(Repeated parameter identifiers, such as -R, -G, and -V may not be specifically 
explained again when they appear below.)

The 1' file was then sampled down to 2' pixels by the command
 grdsample Gebco_SandS_blend_1m.grd=2 -GGebco_SandS_blend_2m.grd=2 -N10800/5400 
-F -Rd -Lx -V
where:
     the output file is Gebco_SandS_blend_2m.grd, with 2-byte binary integer 
pixels
     the -N group specifies 10800 horizontal pixels (specified by the -F option) 
and 5400 vertical pixels
     the -F parameter specifies a pixel, or cell-centered grid
     the -Rd region specification is a shorthand for longitudes +/- 180 degrees 
and latitudes +/- 90 degrees
     the -Lx specifies that the input grid is periodic in the x-direction, so 
there will be no discontinuity at the left or right edge of the output grid.

This basic 2' grid was copied to a file "Gebco_SandS_blend_2mi.raw" to begin the 
assembly process for the whole data base as the bottom layer in the layered 16-
bit grayscale Photoshop file, "ETOPO2v2c<beta>5.psd".

Use of Photoshop as a data assembly tool:

Why Photoshop? The program provides a simple method of masking and overlaying 
binary raster data in an environment of image processing technology. As long as 
the content of individual pixels is not altered other than by replacement, the 
data can be treated the same as pictures with one 16-bit image channel. For 
topographic data, Photoshop's treatment of all binary data as unsigned integers 
can be exploited through the effect of the sign bit, which places all negative 
numbers toward the high end of the brightness scale, and positive values toward 
the bottom. Viewed as an image, topographic data appear with dark land masses 
(values >= zero) and bright oceans. Picking one end of the brightness spectrum 
thus creates an instant land/sea mask which can be handled as a separate layer 
and can be edited to preserve below-sea-level land areas, such as the Dead Sea 
or Death Valley by simply painting their pixels in the mask layer with the 
appropriate mask value. Areas of data with a particular value, such as -500 (the 
no-data flag in GLOBE) can be selected and deleted within a layer leaving a 
transparent void. Overlays of data sets that encompass very large quantities of 
data can be quickly and accurately accomplished without the necessity of 
processing data into the huge ASCII xyz files required by GMT's "grdedit" for 
point-by-point insertion. (Note: Photoshop treats any file with the extension 
".grd" as a specific format for a Photoshop gradient and will not allow the GMT 
data to be read. To avoid this problem, choose file suffixes like ".raw" which 
have no internal special meaning to the application.) Warning: any manipulation 
of signed data in Photoshop other than placing or moving the image elements can 
result in severe data corruption! Smoothing adjacent positive and negative 
numbers in Photoshop will propagate the presence/absence of the sign bit in the 
resulting altered data into unexpected locations. Re-sizing an image of this 
type will have dire consequences. Simple masking or cut-and-paste operations, 
such as performed here, do not change the content of individual pixels in any 
way until layers are flattened. Flattening the image simply leaves the data 
value of each pixel as the value of the topmost visible pixel; transparent 
portions of upper layers allow lower layers' data to be copied through.

IBCAO and GLOBE data were similarly converted to Photoshop 16-bit grayscale 
layers and inserted as layers into the file "ETOPO2v2cbeta5.psd". Each of these 
data layers was downsampled by GMT "grdsample" as follows:

GLOBE:

GLOBE data were read from the CD-ROM data set produced by NGDC as 16 area files 
of 30" pixel data. The raw data files named [A-P]10G were converted to GMT 30" 
grid files by

xyz2grd A10G -GA10G.grd=bs -R-180/-90/50/90 -I30c -F -ZTLhw -N-500 -V
where:
     Region bounds (-R groups) for each grid were defined from the corresponding 
header file information on the CD.    
     the -I30c -F parameters define the output grid as 30 se(c)ond, pixel (-F) 
centered
     the -N-500 parameter sets all data equalling -500 to NaN (Not a Number 
code) in the GMT grid.
     the -ZTLhw group describes the input data as a (h)alf-word (16-bit) binary 
integer raster starting at the (T)op(L)eft corner; the w argument byte-swaps the 
input data to match the requirements of the host (Macintosh G5) computer

and so on, for each source file of 16-bit integer data.

The 16 GMT 30" integer grids were concatenated into increasingly larger areas by 
commands of the type

grdpaste GLOBE_NW2m.grd=2 GLOBE_NE2m.grd=2 -GGLOBE_N2m.grd=2 -V

where:
     the subgrids GLOBE_NW2m.grd and GLOBE_NE2m.grd were in turn constructed by 
grdpaste from smaller [A-P]10g.grd files pasted together into larger 30" grid 
units that covered a quadrant of the globe each. "grdpaste" can join only two 
files at a time into a larger entity, so a repetitive process is required.

To ensure alignment of the output data grid, where appropriate, input data grids 
were first processed into higher-resolution grids at 30" or 1' spacing so that 
the sampling to 2' outputs would not mis-locate the results. This methodology is 
used for the GLOBE, IBCAO, Caspian Sea, Coastal relief Model, and Great Lakes 
data sets.

Each of these large grids was then sampled down to 2-minutes for assembly into 
North and South halves of the final world grid by

grdpaste GLOBE_N2.grd=2 GLOBE_S2.grd=2 -GGLOBE_2.grd=2 -V

Each of the GLOBE_[N-S]2 grids covered a hemisphere. The GMT software could not 
deal with the entire global grid at 30" resolution array size (may have been a 
Mac OS-X or gcc4 compiler implementation restriction on addressing), so the 
split was necessary.

IBCAO:

The IBCAO grid data from NGDC 
(http://www.ngdc.noaa.gov/mgg/bathymetry/arctic/geodata/grd/netcdf/), which is a 
1' grid-registered GMT grid, was resampled by bicubic sampling to 2' pixels by 
the GMT command lines:

grdsample ver1_netcdf_geo.grd -GIBCAOv1_2mI.pix=2 -F -N10800/780 -R-
180/180/64/90 -Lx -V
where:
     the ver1_netcdf_geo.grd is the 1' floating point gridline file was 
downloaded from 
http://www.ngdc.noaa.gov/mgg/bathymetry/arctic/geodata/grd/netcdf/
     the file IBCAOv1_2mI.pix=2 is the output grid of 2-byte integer 2' pixels 
extracted
     the region and pixel count parameters -R and -N are as defined above.
     Sampling by bicubic spline is the default method for this program; nearest-
neighbor sampling can also be selected by the -Q parameter (not used here)

This made a grid of 2' pixels in a standard GMT format that can be imported as 
16-bit "raw" format by Photoshop with a 892-byte header that can be specified to 
be ignored on the "Open" dialog in Photoshop. Photoshop "raw" files are assumed 
to be uncompressed binary pixel rasters with no compression or encoding. This 
layer was copied to the clipboard and pasted in at the top of the 
"ETOPO2v2cbeta5.psd" image.

The 10800x5400-pixel Photoshop file was then flattened, with the top-to-bottom 
layer order being IBCAO-GLOBE-SandS. This made the precedence of data favor 
IBCAO for all data at 64 degrees and farther north; GLOBE topography overlays 
the rest of the (GTOPO30) land data from SandS. Ocean bathymetry from 80 South 
to 64 North came through from the SandS layer. The result was output, with the 
header size set to zero in the "Save" dialog in Photoshop, to form the basic 
grid file "ETOPO2v2cbeta5.raw"

The "ETOPO2v2cbeta5.raw" file was then input to a GMT 2-byte integer cell-
centered grid by the GMT command: 

xyz2grd ETOPO2v2cbeta5.raw -Rd -I2m -ZTLh -GETOPO2v2c.5.grd=2 -F -V

Subsequent data sets were inserted into ETOPO2v2c.5.grd via GMT software as 
follows:

Bathymetry data for the Caspian Sea were obtained as digitized contours from the 
Caspian Environment Programme website at 
http://www.caspianenvironment.org/dim/menu5.htm. The file "isobat_ascii.zip" was 
downloaded and unpacked into multiple text files. Each of these was converted in 
an Excel spread sheet from 2-column latitude-longitude data streams for each 
contour to xyz data with the third (z) column containing the contour value, that 
could be read by GMT's "surface" program. All of the xyz contour files were 
concatenated into a single file (Caspian.blm) and converted to a 30" grid 
through GMT "surface" and resampled to the needed 2' spacing with GMT 
"grdsample":

surface Caspian.blm -GCaspian30s.grd -R45.5/56/36.5/47.5 -I30c -V

Output from "surface" is always grid-registered, rather than cell-centered.

The resulting 30" grid was resampled to a 2' floating-point grid by:

grdsample Caspian30s.grd -GCaspian2mF.grd -F -N315/330 -R45.5/56/36.5/47.5 -V

The sampling to 2' used a bicubic spline to fill the cells, using the default 
parameter that requires all 4 nearest source nodes to have data for the result 
to be included in the output grid. Resulting depth values start at zero for the 
shoreline. relative to the local Caspian Sea surface datum, to be adjusted 
below. The grid was then masked to the sea boundary according to the GMT 
maximum-resolution coastline file:

grdlandmask -GCaspian_maskF.grd -R45.5/56/36.5/47.5 -I2m -Df -N1/NaN/1/NaN/NaN -
V -F
where:
     the -N group defines data not included within a "sea" or "lake" boundary to 
be excluded as Not-a-Number (NaN) in the resulting grid (only first and third 
arguments set to 1).
     the -Df parameter specifies the "full-resolution" GMT coastline file, in 
which the Caspian Sea is flagged as a "lake".

The resulting data were biased by the Caspian Sea datum of -29m so they fit the 
below-sea-level elevations of the surrounding basin. After this conversion to a 
new grid, these data were converted to xyz ASCII data:

grd2xyz Caspian_2mFmaskedI.grd=2 -S -R45.5/56/36.5/47.5 -V > CaspianBathy.xyz
where
     the -S parameter sets output to actual data only, without NaN points.
and edited into the intermediate ETOPO2v2c grid:

grdedit ETOPO2v2c.5.grd=2 -NCaspianBathy.xyz -V
where:
     the -N specifies the input (N)ew data file

Insertion of these data superseded existing grid points closest to each xyz data 
point location.

For coastal USA waters, the NGDC Coastal Relief Model (CRM) 3" grids for water 
only were downloaded from CRM Volumes 1-10 via GEODAS as 15" subsamples and 
resampled with "grdsample", converted to 2' xyz data and inserted over their 
corresponding positions in the 2' data, e.g.:

grdsample Vol10.grd -GVol10-2m.grd=2 -R-161/-153/18/23 -N240/150 -F -V

grd2xyz -S Vol10-2m.grd=2 -V  > Vol10-2m.xyz

grdedit ETOPO2v2c.5.grd=2 -NVol10-2m.xyz -V

and so on for data from CRM Volumes 1-9 as well.

Great Lakes bathymetry data were added by conversion from 3" NGDC source grids 
for Lakes Erie, Ontario, Huron, and Michigan; preliminary 30" source data were 
used for Lake Superior. Each lake's water data were isolated and adjusted to its 
mean level datum, obtained from GLERL (NOAA Great Lakes Environmental Research 
Laboratory). Resulting xyz grids were generated and inserted as above. 


The current (30 May 2006) version of the data base is ETOPO2v2c, available on the
CD-ROM as floating-point GMT (.grd=bf) and MSB-first, or Big-Endian (.raw) 
raster grids, and ArcMap (.flt and .hdr) files. A Miller-Projection, color, 
shaded-relief image of the data is also included as an Acrobat (.pdf) file. The
GMT parameters describing the grid are "-R-180/80/-90/90 -I2m -F".


Peter W. Sloss, 25 July 2006
ngdc.info@noaa.gov

Other Forms:

  ETOPO2v2g is in a grid-centered registration (pixel registered); the cell
  boundaries are lines of odd minutes of latitude and longitude, and they are
  centered on intersections of lines of even minutes of latitude and longitude.

Availability:

ETOPO2v2 is available in its entirety on DVD:
http://www.ngdc.noaa.gov/mgg/fliers/06mgg01.html