Content uploaded by Dean B. Gesch
Author content
All content in this area was uploaded by Dean B. Gesch on Jul 23, 2014
Content may be subject to copyright.
U.S. Department of the Interior
U.S. Geological Survey
Open-File Report 2011–1073
Global Multi-resolution Terrain Elevation Data 2010
(GMTED2010)
Cover illustration: GMTED2010 dataset.
Global Multi-resolution Terrain Elevation
Data 2010 (GMTED2010)
By Jeffrey J. Danielson and Dean B. Gesch
Open-File Report 2011–1073
U.S. Department of the Interior
U.S. Geological Survey
U.S. Department of the Interior
KEN SALAZAR, Secretary
U.S. Geological Survey
Marcia K. McNutt, Director
U.S. Geological Survey, Reston, Virginia: 2011
For more information on the USGS—the Federal source for science about the Earth, its natural and living
resources, natural hazards, and the environment, visit http://www.usgs.gov or call 1–888–ASK–USGS.
For an overview of USGS information products, including maps, imagery, and publications,
visit http://www.usgs.gov/pubprod
Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the
U.S. Government.
Although this report is in the public domain, permission must be secured from the individual copyright owners to
reproduce any copyrighted materials contained within this report.
Suggested citation:
Danielson, J.J., and Gesch, D.B., 2011, Global multi-resolution terrain elevation data 2010 (GMTED2010): U.S. Geo-
logical Survey Open-File Report 2011–1073, 26 p.
iii
Contents
Introduction.....................................................................................................................................................1
Existing GTOPO30 Elevation Model ............................................................................................................1
GMTED2010 Dataset Characteristics .........................................................................................................2
Input Data Sources ........................................................................................................................................2
Data Preprocessing .......................................................................................................................................5
Generalization.................................................................................................................................................6
Mosaicking......................................................................................................................................................7
Pixel Alignment and Grid Coordinates .......................................................................................................7
Output Data Products ....................................................................................................................................9
Data Formats.................................................................................................................................................12
Generalized Elevation Products – Seamless Global Coverage (GRID) ......................................12
Generalized Elevation Products – Tile-Based (GeoTIFF Image Format) ....................................12
Spatially Referenced Metadata (ESRI Vector Shapefile Format) ...............................................12
Accuracy Assessment ................................................................................................................................12
Raster-Based Assessment ................................................................................................................12
Absolute Vertical Accuracy Assessment .......................................................................................12
Limitations and Caveats ..............................................................................................................................21
Summary........................................................................................................................................................21
References Cited..........................................................................................................................................22
Appendix........................................................................................................................................................25
Figures
1. Map showing GTOPO30 elevation model .................................................................................1
2. Map showing GTOPO30 elevation sources ..............................................................................2
3. Map showing SRTM DTED®2 (void-filled) 1-arc-second coverage .....................................3
4. Map showing 15-arc-second SPOT 5 Reference3D Africa coverage .................................4
5. Diagram showing aggregate example using the maximum value (3 x 3 processing
window) ..........................................................................................................................................6
6. Diagram showing standard deviation example using blockstd routine (3 x 3 processing
window) ..........................................................................................................................................7
7. Diagram showing ArcGIS mosaic – blend method .................................................................8
8. Diagram showing “Pixel center” referencing of full resolution 1-arc-second SRTM
data .................................................................................................................................................8
9. Diagram showing spatial nesting of GMTED2010 pixels .............................................................8
10. Diagram showing the GMTED2010 layer extents (minimum and maximum latitude and
longitude) are a result of the coordinate system inherited from the 1-arc-second SRTM
data ................................................................................................................................................10
11. Graphic showing comparison of the existing GTOPO30 and new GMTED2010 30-arc-
second mean elevation product ...............................................................................................10
12. Map showing GMTED2010 spatially referenced metadata with void polygons having
a dark dense appearance .........................................................................................................11
iv
13. Graph showing GMTED2010 mean product: 30-arc-second error statistics by source
(points in red are the number of control points used for each source) .............................19
14. Graph showing GMTED2010 mean product: 15-arc-second error statistics by
source (points in red are the number of control points used for each source) ...............19
15. Graph showing GMTED2010 mean product: 7.5-arc-second error statistics by
source (points in red are the number of control points used for each source) ...............20
16. Graph showing GMTED2010: percent land area by source .................................................20
17. Graph showing GMTED2010: global aggregated product accuracy ..................................20
Tables
1. Global land area percentage by source ...................................................................................5
2a. Input source data characteristics .............................................................................................5
2b. Input source data characteristics .............................................................................................6
3. Geographic extents, resolutions, and raster grid dimensions of GMTED2010 product
layers ..............................................................................................................................................9
4a. Spatially referenced metadata attribute data dictionary .....................................................11
4b. Example values - GMTED2010 spatially referenced metadata ...........................................11
5. GMTED2010 global accuracy assessment: raster-based comparison: GMTED2010
30-arc-second products minus GTOPO30 (meters) ..............................................................12
6. Removal of outliers beyond three standard deviations from the mean difference
between NGA control point dataset and the GMTED2010 systematic subsample
product .........................................................................................................................................13
7. GMTED2010 absolute accuracy assessment: aggregated global data evaluation:
GMTED2010 products and GTOPO30 minus NGA control points (meters) ........................14
8a. GMTED2010 absolute accuracy assessment by source: SRTM DTED®2: GMTED2010
products minus NGA control points (meters) ................................................................................15
8b. GMTED2010 absolute accuracy assessment by source: CDED1: GMTED2010 products
minus NGA control points (meters) .................................................................................................15
8c. GMTED2010 absolute accuracy assessment by source: CDED3: GMTED2010 products
minus NGA control points (meters) .................................................................................................16
8d. GMTED2010 absolute accuracy assessment by source: DTED®1: GMTED2010 products
minus NGA control points (meters) .................................................................................................16
8e. GMTED2010 absolute accuracy assessment by source: GTOPO30 fill data: GMTED2010
products minus NGA control points (meters) ................................................................................17
8f. GMTED2010 absolute accuracy assessment by source: NED: GMTED2010 products
minus NGA control points (meters) .................................................................................................17
8g. GMTED2010 absolute accuracy assessment by source: NED-Alaska: GMTED2010
products minus NGA control points (meters) ................................................................................18
8h. GMTED2010 absolute accuracy assessment by source: 15-Arc-second SPOT5 Refer-
ence3D: GMTED2010 products minus NGA control points (meters) ..........................................18
8i. GMTED2010 absolute accuracy assessment by source: Greenland: GMTED2010
products minus NGA control points (meters) ................................................................................19
Introduction
In 1996, the U.S. Geological Survey (USGS) developed
a global topographic elevation model designated as GTOPO30
at a horizontal resolution of 30 arc-seconds for the entire
Earth. Because no single source of topographic information
covered the entire land surface, GTOPO30 was derived from
eight raster and vector sources that included a substantial
amount of U.S. Defense Mapping Agency data. The quality
of the elevation data in GTOPO30 varies widely; there are
no spatially-referenced metadata, and the major topographic
features such as ridgelines and valleys are not well repre-
sented. Despite its coarse resolution and limited attributes,
GTOPO30 has been widely used for a variety of hydrological,
climatological, and geomorphological applications as well as
military applications, where a regional, continental, or global
scale topographic model is required. These applications have
ranged from delineating drainage networks and watersheds to
using digital elevation data for the extraction of topographic
structure and three-dimensional (3D) visualization exercises
(Jenson and Domingue, 1988; Verdin and Greenlee, 1996;
Lehner and others, 2008). Many of the fundamental geophysi-
cal processes active at the Earth’s surface are controlled or
strongly inuenced by topography, thus the critical need for
high-quality terrain data (Gesch, 1994). U.S. Department of
Defense requirements for mission planning, geographic regis-
tration of remotely sensed imagery, terrain visualization, and
map production are similarly dependent on
global topographic data.
Since the time GTOPO30 was com-
pleted, the availability of higher-quality
elevation data over large geographic
areas has improved markedly. New data
sources include global Digital Terrain
Elevation Data (DTED®) from the Shuttle
Radar Topography Mission (SRTM),
Canadian elevation data, and data from
the Ice, Cloud, and land Elevation Satel-
lite (ICESat). Given the widespread use
of GTOPO30 and the equivalent 30-arc-
second DTED® level 0, the USGS and the
National Geospatial-Intelligence Agency
(NGA) have collaborated to produce an
enhanced replacement for GTOPO30, the
Global Multi-resolution Terrain Elevation Data 2010
(GMTED2010)
By Jeffrey J. Danielson and Dean B. Gesch
Global Land One-km Base Elevation (GLOBE) model and
other comparable 30-arc-second-resolution global models,
using the best available data. The new model is called the
Global Multi-resolution Terrain Elevation Data 2010, or
GMTED2010 for short. This suite of products at three differ-
ent resolutions (approximately 1,000, 500, and 250 meters) is
designed to support many applications directly by providing
users with generic products (for example, maximum, mini-
mum, and median elevations) that have been derived directly
from the raw input data that would not be available to the
general user or would be very costly and time-consuming
to produce for individual applications. The source of all the
elevation data is captured in metadata for reference purposes.
It is also hoped that as better data become available in the
future, the GMTED2010 model will be updated.
Existing GTOPO30 Elevation Model
GTOPO30, a widely used global elevation model, was
produced by the USGS and became available in 1996 (Gesch
and others, 1999). GTOPO30 provides elevations for the entire
global land surface on a grid every 30 arc-seconds of latitude
and longitude, which is about 1-kilometer spacing at the equa-
tor (g. 1).
At the time GTOPO30 was developed, and even today,
no one source of topographic information covered the entire
Figure 1. GTOPO30 elevation model.
2 Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010)
land surface. GTOPO30 was derived from eight raster and
vector sources of varying degrees of quality with process-
ing techniques differing from continent to continent (Gesch
and Larson, 1998; Gesch and others, 1999) (g. 2). Since
GTOPO30 was completed, the availability of high-quality
elevation data over large areas has improved markedly. These
new data sources provide a substantial improvement over the
inputs to GTOPO30 with respect to consistent coverage, scale,
quality, and vertical accuracy.
GMTED2010 Dataset Characteristics
The USGS and the NGA have collaborated on the devel-
opment of a notably enhanced global elevation model named
the GMTED2010 that replaces GTOPO30 as the elevation
dataset of choice for global and continental scale applications.
The new model has been generated at three separate resolu-
tions (horizontal post spacings) of 30 arc-seconds (about
1 kilometer), 15 arc-seconds (about 500 meters), and 7.5 arc-
seconds (about 250 meters). This new product suite provides
global coverage of all land areas from lat 84°N to 56°S for
most products, and coverage from 84°N to 90°S for several
products. Some areas, namely Greenland and Antarctica, do
not have data available at the 15- and 7.5-arc-second resolu-
tions because the input source data do not support that level
of detail. An additional advantage of the new multi-resolution
global model over GTOPO30 is that seven new raster eleva-
tion products are available at each resolution. The new
elevation products have been produced using the following
aggregation methods: minimum elevation, maximum eleva-
tion, mean elevation, median elevation, standard deviation of
elevation, systematic subsample, and breakline emphasis. The
systematic subsample product is dened using a nearest neigh-
bor resampling function, whereby an actual elevation value is
extracted from the input source at the center of a processing
window. Most vertical heights in GMTED2010 are referenced
to the Earth Gravitational Model 1996 (EGM 96) geoid (NGA,
2010). In addition to the elevation products, detailed spa-
tially referenced metadata containing attribute elds such as
coordinates, projection information, and raw source elevation
statistics have been generated on a tile-by-tile basis for all the
input datasets that constitute the global elevation model.
Input Data Sources
GMTED2010 is based on data derived from 11 raster-
based elevation sources. The primary source dataset for
GMTED2010 is NGA’s SRTM Digital Terrain Elevation
Data (DTED®2, http://www2.jpl.nasa.gov/srtm/) (void-lled)
1-arc-second data. For the geographic areas outside the SRTM
coverage area and to ll in remaining holes in the SRTM data,
the following sources were used: (1) non-SRTM DTED®,
(2) Canadian Digital Elevation Data (CDED) at two resolu-
tions, (3) Satellite Pour l’Observation de la Terre (SPOT 5)
Reference3D, (4) National Elevation Dataset (NED) for the
continental United States and Alaska, (5) GEODATA 9 second
digital elevation model (DEM) for Australia, (6) an Antarctica
satellite radar and laser altimeter DEM, and (7) a Greenland
satellite radar altimeter DEM. Each is described below.
The SRTM data cover 80 percent of the Earth’s land sur-
face (all latitudes between 60°N and 56°S) (g. 3) and provide
a substantial upgrade over the primary source datasets used for
GTOPO30, the older 3-arc-second DTED®1, and Digital Chart
of the World (DCW) 1:1,000,000-scale cartographic data
produced by NGA. The original SRTM data processing and
editing is documented in Farr and others (2007), and Slater
and others (2006). The void-lled SRTM data are a revised
version of the original NGA dataset that is not currently
Figure 2. GTOPO30 elevation sources.
EXPLANATION
Digital Terrain Elevation Data
Digital Chart of the World
USGS Digital Elevation Models
Army Map Service Maps
International Map of the World
Peru Map
New Zealand DEM
Antarctic Digital Database
GTOPO30 Source Data
Input Data Sources 3
publicly available. The void-lled version includes additional
spike/well removal using a threshold of 60 meters (instead of
the original 100 meters) with respect to the surrounding terrain
and the detection and removal of phase unwrapping errors that
were remnants of the original raw radar data processing. After
detecting these artifacts in the data, the corresponding eleva-
tion posts were voided out and then systematically replaced
with an alternate source of elevation data primarily from
non-SRTM DTED®, NED, and SPOT 5. In places where no
acceptable alternate source data were available and where the
size of the void and the surrounding terrain were appropriate,
interpolation was used to ll the void. Although most of the
data voids in the 1-arc-second SRTM data have been lled by
NGA, some residual voids remain where suitable source data
at the required spatial resolution were not available and no
interpolation was done. For these areas, GMTED2010 produc-
tion included lling the residual voids in the SRTM DTED®2
dataset with the vertical heights referenced to the EGM96
geoid.
Canadian Digital Elevation Data (CDED, http://
www.geobase.ca/doc/specs/pdf/GeoBase_product_specs_
CDED1_3_0.pdf) consists of an ordered array of ground or
reective surface elevations recorded in meters at regularly
spaced intervals of 0.75 and 3 arc-seconds. The digital source
data for CDED are extracted from the hypsographic and
hydrographic elements of the National Topographic Data Base
(NTDB) at scales of 1:50,000 and 1:250,000, the Geospatial
Database (GDB), various scaled positional data acquired by
the provinces and territories, or remotely sensed imagery. Ver-
tical heights in CDED are referenced to the Canadian Vertical
Geodetic Datum of 1928 (CVGD28). CDED was used as the
source for areas north of lat 60°N in Canada and is available
from GeoBase (http://www.geobase.ca/geobase/en/data/cded/
index.html) Canada.
Non-SRTM DTED® Level 1 is a raster topographic
database of terrain elevation values with post spacings every
3 arc-seconds (approximately 100 meters). The information
content is approximately equivalent to the contour information
represented on a 1:250,000 scale map. Non-SRTM DTED®
Level 1 is photogrammetrically-derived and produced by
NGA. Vertical heights in Non-SRTM DTED®1 are referenced
to Mean Sea Level (MSL). The Non-SRTM DTED®1 was
used as the source for areas north of lat 60°N in Eurasia and
for void-lling SRTM DTED®2. SRTM DTED®2 was used for
areas below lat 60°N.
SPOT 5 Reference3D (http://www.spot.com/automne_
modules_les/standard/public/p453_e66fdc8d9baeb629a19
beb53be67339dReference3D-Product_descriptionv5-2.pdf)
is a uniform grid of terrain elevation values that are obtained
through automatic correlation of SPOT High-Resolution
Stereoscopic (HRS) image pairs (Bouillon and others, 2006).
SPOT 5 Reference3D is co-produced by Spot Image (http://
www.spot.com/?countryCode=US&languageCode=) and the
Institut Geographique National (IGN, http://www.ign.fr/),
France’s national survey and mapping agency. Spot Image and
IGN provided their global Reference3D collection at no cost
to the USGS at a generalized 15-arc-second resolution for the
sole purpose of SRTM void-lling. The method selected to
generalize the Reference3D from its native resolution to the
15-arc-second resolution is based on selecting a single eleva-
tion value from the Reference3D at the center of the 15-arc-
second output processing window. This method is comparable
to a nearest neighbor resampling technique. The 15-arc-second
SPOT 5 Reference3D dataset represents the only contribu-
tion to this project by a private commercial company. Vertical
heights in SPOT5 Reference3D are referenced to the EGM96
geoid. The 15-arc-second SPOT 5 Reference3D was used for
lling SRTM voids in Africa (g. 4), Central America, Asia,
and Australia.
The National Elevation Dataset (NED, http://ned.usgs.
gov/) is a seamless dataset with the best available raster eleva-
tion data of the conterminous United States, Alaska, Hawaii,
and territorial islands. The NED provides elevation data in a
consistent datum, elevation unit, and projection. NED data are
available nationally (except for Alaska) at resolutions of 1 arc-
second (about 30 meters) and 1/3 arc-second (about
Figure 3. SRTM DTED®2 (void-filled) 1-arc-second coverage map.
4 Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010)
10 meters), and in limited areas at 1/9 arc-second (about
3 meters). Vertical heights in NED are referenced to the North
American Vertical Datum of 1988 (NAVD 88). In most of
Alaska, only lower resolution source data are available. As
a result, most NED data for Alaska are at the 2-arc-second
(about 60-meter) grid spacing (Gesch, 2007). Vertical heights
in the 2-arc-second NED for Alaska are referenced to the
National Geodetic Vertical Datum of 1929 (NGVD 29). NED
is produced by the USGS (http://www.usgs.gov/) and was used
as a void-ll source for the SRTM tiles over the conterminous
United States and for all areas in Alaska north of lat 60°N.
The GEODATA 9 Second Digital Elevation Model
(DEM, http://www.ga.gov.au/meta/ANZCW0703011541.html)
Version 2 is a grid of ground level elevation points covering
Australia with a grid spacing of 9 seconds in longitude and
latitude (approximately 250 meters) in the Geocentric Datum
of Australia 1994 (GDA 94) coordinate system. GEODATA
9 Version 2 is improved over the previous version by includ-
ing the national trigonometric data points in the source data
and by revising the gridding procedure to model high points
and breaklines more precisely. Abrupt changes in landform
have also been modeled in Version 2 by incorporating cliff
line data in selected areas. Vector-based data included in the
gridding include 1:100,000 scale topographic spot elevation
heights, 1:250,000 scale streams and contours, and national
trigonometric data points supplied from the National Geo-
detic Data Base (NGDB). Vertical heights in the GEODATA 9
Second Digital Elevation Model (DEM) are referenced to the
Australian Height Datum (AHD). This dataset is produced by
Geoscience Australia (http://www.ga.gov.au/) and was used as
a void-ll source for the SRTM tiles over Australia.
A new elevation model for Antarctica (Bamber and oth-
ers, 2008) developed at the University of Bristol (http://www.
bris.ac.uk/) is distributed by the National Snow and Ice Data
Center (NSIDC, http://nsidc.org/). The new Antarctica model
is a grid regularly spaced at 1-kilometer intervals and has been
created from combined European Radar Satellite (ERS-1)
radar data and ICESat laser satellite altimetry. The ERS-1 data
are from two long repeat cycles of 168 days initiated in March
1994, and the ICESat Geoscience Laser Altimeter System
(GLAS) data (Zwally and others, 2007) are from February
20, 2003, through March 21, 2008. Vertical heights in the
Antarctica DEM are referenced to the World Geodetic System
1984 (WGS 84) ellipsoid but were converted to the EGM96
geoid for use in GMTED2010.
A gridded DEM at a 1-kilometer post spacing developed
from ERS-1 and Geosat satellite radar altimetry (Bamber
and others, 2001) provides the source data for the Greenland
portion of GMTED2010. Where the ERS-1 and Geosat radar
altimetry were lacking sufcient spatial coverage in bare rock
regions, stereophotogrammetric datasets, synthetic aperture
radar, and digitized cartographic maps were used. An accu-
racy assessment of the Greenland DEM was completed using
airborne laser altimetry that had an accuracy between 10 and
12 centimeters. The mean vertical accuracy of the Greenland
DEM was determined to be –0.33 ±6.97 meters over the entire
ice sheet. Depending on the input data source, the accuracy
over bare rock regions ranges from 20 to 200 meters (Bamber
Figure 4. 15-arc-second SPOT 5 Reference3D Africa coverage map.
EXPLANATION
Data Preprocessing 5
and others, 2001). Vertical heights in the Greenland DEM are
referenced to the WGS 84 ellipsoid but were converted to the
EGM96 Geoid for use in GMTED2010.
The percentage of the global land surface area derived
from each GMTED2010 data source is displayed in table 1.
GTOPO30 was used only as a source of last choice for lling
residual voids in SRTM data.
Data Preprocessing
Data characteristics such as the projection system, coor-
dinate units, and horizontal and vertical datum vary among
the input data sources (tables 2a and 2b). These input data
characteristics (except for vertical datum) were standardized
to a consistent set of parameters in order to create a seamless
global elevation dataset. Every input dataset was ingested on
a tile-by-tile (1° x 1°) basis and transformed to the geographic
WGS84 horizontal coordinate system with their respective
horizontal units converted to decimal degrees and vertical
units changed to integer meters. The Project Raster tool within
ArcGIS 9.3 was used to carry out the data transformation to
WGS84 using a bilinear resampling option. Vertical datum dif-
ferences between the input data sources were not transformed
but captured in the spatially referenced metadata, except for
Greenland and Antarctica, where NGA transformed the verti-
cal datum.
Voids in the SRTM data were lled using the Delta Sur-
face Fill (DSF) method developed by NGA (Grohman and oth-
ers, 2006). The DSF method replaces the void with ll source
posts that are adjusted to the SRTM values found at the void
interface. This process causes the ll to more closely follow
the trend of the original SRTM surface while retaining
the useful characteristics from the source ll data. A
total of 1,573 1 x 1-degree SRTM tiles with partial missing
data were lled using most of the source datasets listed in
table 2a. The accompanying spatially referenced metadata
document the source used to ll a particular area. There were
three main causes for voids in the SRTM data: a few patches
of land in North America were missed because the radar sen-
sor did not collect data during 10 orbits of the mission, steep
slopes caused shadow layover effects, and certain areas with
sandy soils (for example, northern Africa) caused poor radar
returns (Farr and others, 2007).
NGA received two les for the Greenland DEM (Bamber
and others, 2001): (1) orthometric heights (H) and (2) cor-
responding geoid heights (N) at 30-arc-second resolution. File
(1) had geographic boundaries of lat 83.75° to 59.4°N,
long 74° to 11°W, and le (2) had boundaries of lat 84° to
59.5°N, long 75° to 10°W. Although there was a slight offset
between the les, they were aligned and both have a grid spac-
ing of 36 arc-seconds in latitude and 90 arc-seconds in longi-
tude. After conrming the sign convention used in the geoid
height le (2), NGA recovered the WGS 84 ellipsoid heights
(h) by adding the geoid height to the orthometric height: H +
Table 1. Global land area percentage by source.
[SRTM, Shuttle Radar Topography Mission; DTED®, Digital Terrain
Elevation Data; DEM, digital elevation model; CDED, Canadian Digital
Elevation Data; NED, National Elevation Dataset; SPOT, Satellite Pour
l’Observation de la Terre; GTOPO30, Global 30-Arc-Second Elevation
Dataset]
Dataset Percent land
area
SRTM (DTED® 2) 69.92
Antarctica satellite radar and laser altimeter DEM 13.80
DTED® 1 8.71
CDED3 2.26
CDED1 2.24
Greenland satellite radar altimeter DEM 1.79
NED – Alaska 1.01
15-arc-second SPOT 5 Reference3D 0.16
GTOPO30 0.09
NED 0.01
GEODATA 9 second DEM version 2 *
*0.0004 percent
Table 2a. Input source data characteristics.
[SRTM, Shuttle Radar Topography Mission; DTED®, Digital Terrain Elevation Data; NGA, National Geospatial-Intelligence Agency; WGS 84, World
Geodetic System 1984; CDED, Canadian Digital Elevation Data; NAD 83, North American Datum of 1983; SPOT, Satellite Pour l’Observation de la
Terre; IGN, Institut Geographique National; NED, National Elevation Dataset; USGS, U. S. Geological Survey; DEM, digital elevation model; GDA 94,
Geocentric Datum of Australia 1994; GTOPO30, Global 30-Arc-Second Elevation Dataset]
Dataset Source organization Resolution Horizontal unit Horizontal
datum
SRTM DTED® 2 NGA 1 Arc-second WGS 84
DTED® 1 NGA 3 Arc-second WGS 84
CDED1 GeoBase - Canada 0.75 Arc-second NAD 83
CDED3 GeoBase - Canada 3 Arc-second NAD 83
15-arc-second SPOT 5 Reference3D Spot Image / IGN 0.00416666 Decimal degree WGS 84
NED USGS 0.00027777 Decimal degree NAD 83
NED – Alaska USGS 0.00055555 Decimal degree NAD 83
GEODATA 9 second DEM version 2 Geoscience Australia 0.0025 Decimal degree GDA 94
Greenland satellite radar altimeter DEM University of Bristol 1,000 Meter WGS 84
Antarctica satellite radar and laser altimeter DEM University of Bristol 1,000 Meter WGS 84
GTOPO30 USGS 0.00833333 Decimal degree WGS 84
6 Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010)
N = h. All values not over Greenland were set to 0.0 (where
–0.1 was coded in the orthometric height le (1)). This dataset
le was then bilinearly interpolated to the inner 30-arc-second
by 30-arc-second grid points. Values correspond to the center
of the 30-arc-second equi-angular grid cell. The EGM96 geoid
height was then subtracted from the WGS 84 ellipsoid height
to produce the EGM96 orthometric height at each 30-arc-
second by 30-arc-second grid point. The resulting le was then
“land-agged” using NGA land agging software in conjunc-
tion with the World Vector Shoreline continental outline le.
Negative values were set to 0.0. All land values not in Green-
land were eliminated and all areas outside Greenland were set
to 0.0. Additional error was introduced in the coastline through
the interpolation process.
NGA received one le for the Antarctica DEM (Bamber
and others, 2008). The DEM is on a 5,601 x 5,601 1-kilometer
grid with the center at the South Pole and on a polar stereo
projection with the standard parallel at lat 71°S. Elevations are
with respect to the WGS 84 ellipsoid. NGA used the GEO-
TRANS version 2.3 software to convert the data from polar
stereographic to geodetic coordinates. The geoid undulations
were computed in meters for the entire Antarctica dataset
using EGM96 software with geopotential/correction coef-
cients. The geoid undulations were then subtracted from the
ellipsoid heights to produce orthometric heights.
Generalization
Data processing was accomplished by developing
workows in Python 2.5.1 and accessing Environmental
System Research Institute’s (ESRI) ArcGIS 9.3.1 geoprocess-
ing framework to perform raster and vector spatial analysis
operations. The Geospatial Data Abstraction Library (GDAL),
an open source image processing package, was used for the
mosaic compilation of each continental area. To more ef-
ciently handle the numerous input datasets and to standardize
the proper sequence of processing steps, the production
procedures were automated to a great extent by using preset
parameter values, scripted routines, and consistent naming
conventions for input and output data les.
The generalization, or aggregation, approach produces
reduced resolution data that represent the minimum, maxi-
mum, mean, and median of the full resolution source eleva-
tions within the aggregated output cell. The statistical-based
products were generated using the Aggregate function within
ArcGIS. The Aggregate function resamples an input raster
grid to a coarser resolution based on a specied aggregation
strategy (Minimum, Maximum, Mean, or Median)
(g. 5). The pixel resolution (horizontal resolution, post-
spacing) of the input raster grid is multiplied by the cell factor,
which corresponds to the desired pixel resolution of the output
raster grid.
In addition, a systematic subsampling of the full resolu-
tion source data was used to produce a reduced resolution
version at each of the output grid spacings. The systematic
subsample product was computed using the Resample func-
tion in ArcGIS with the nearest neighbor option. The standard
deviation product was generated using a combination of two
functions in ArcGIS. A Blockstd function was rst applied that
partitions the input raster grid into blocks, nds the standard
Figure 5. Aggregate example using the maximum value (3 x 3
processing window).
Table 2b. Input source data characteristics.
[SRTM, Shuttle Radar Topography Mission; DTED®, Digital Terrain Elevation Data; EGM96, Earth Gravitational Model 1996; MSL, Mean Sea Level;
CDED, Canadian Digital Elevation Data; CVGD28, Canadian Vertical Geodetic Datum of 1928; SPOT, Satellite Pour l’Observation de la Terre; NED,
National Elevation Dataset; NAVD 88, North American Vertical Datum of 1988; NGVD 29, National Geodetic Vertical Datum of 1929; DEM, digital
elevation model; AHD, Australian Height Datum; WGS 84, World Geodetic System 1984; GTOPO30, Global 30-Arc-Second Elevation Dataset]
Dataset Projection system Vertical unit Vertical datum Surface type
SRTM DTED® 2 Geographic Integer meter EGM 96 Geoid Reective
DTED® 1 Geographic Integer meter MSL Bare-Earth
CDED1 Geographic Integer meter CVGD28 Reective
CDED3 Geographic Integer meter CVGD28 Reective
15-arc-second SPOT 5 Reference3D Geographic Integer meter EGM96 Geoid Reective
NED Geographic Decimal meter NAVD 88 Bare-earth
NED – Alaska Geographic Decimal meter NGVD 29 Bare-earth
GEODATA 9 second DEM version 2 Geographic Integer meter AHD Bare-earth
Greenland satellite radar altimeter DEM Polar stereographic Integer meter WGS 84 Ellipsoid Reective
Antarctica satellite radar and laser altimeter DEM Polar stereographic Integer meter WGS 84 Ellipsoid Reective
GTOPO30 Geographic Integer meter MSL Bare-earth
Pixel Alignment and Grid Coordinates 7
deviation for the specied posts dened by the neighborhood
blocks, and sends the computed standard deviation to the post
locations in the corresponding blocks on the output raster
grid (g. 6). The Blockstd output was then generalized to the
desired output resolution of 30, 15, or 7.5 arc-seconds using a
nearest neighbor resampling.
Another approach called “breakline emphasis” was used
to produce reduced resolution products that maintain stream
(channel) and ridge (divide) characteristics as delineated in the
full resolution source data (Gesch, 1999). Breakline emphasis
maintains the critical topographic features within the land-
scape by maintaining any minimum elevation or maximum
elevation value on a breakline that passes within the specied
analysis window. Remaining elevation values are generalized
using the median statistic. The breakline emphasis methodol-
ogy can be summarized into three steps:
1. Topographic breaklines (ridges and streams) are
extracted from the full resolution DEM and then used
to guide selection of generalized values.
2. Full resolution streams are automatically thresholded,
which enables easy extraction of the level one
through ve Strahler stream orders.
3. Full resolution ridges are extracted by selecting the
ow accumulation values that are equal to zero.
Using focal and block image processing functions,
ridges are thinned so that only critical divides are
maintained.
The breakline emphasis product is especially useful
for the generation of hydrologic derivatives or distributed
hydrologic modeling applications conducted over large areas
(Danielson and Gesch, 2008). One practical way to test the
operational effectiveness of the breakline emphasis algorithm
is to generate watersheds from the generalized product and
to compare the result against watersheds derived from a full
resolution data source. This was demonstrated by Danielson
and Gesch (2008) for the Allegheny and James River basins
in the United States. The authors concluded that for these two
very different physiographic regions the breakline emphasis
approach maintained a 97.3 percent spatial agreement between
the watersheds derived from the reduced resolution elevation
model and watersheds derived from the full resolution source
elevation model.
Mosaicking
Generalization was completed on the input source data
on a tile-by-tile basis. Input tiles were then mosaicked into
dataset level mosaics. Depending on the number of data
sources over a particular continental area, different source
dataset level mosaics were then mosaicked into continental
area mosaics. The continental areas were then mosaicked into
the nal global mosaics at each of the three resolutions for all
seven products. Global mosaicking was accomplished using
the Mosaic function within ArcGIS. The function creates one
grid from two or more adjacent grids and makes a smooth
transition over the overlapping areas of the neighboring grids
using the blend method (http://resources.esri.com/help/9.3/
arcgisengine/com_cpp/gp_toolref/data_management_tools/
mosaic_methods.htm). This method uses a distance-weighted
algorithm to determine the value of overlapping elevation
posts. The output elevation value of the overlapping areas
will be a blend of elevation values that overlap. This blend
value is based on an algorithm that is weight based and is
dependent on the distance from the post to the edge within the
overlapping area. A diagram that shows two overlapping raster
datasets is shown in gure 7. The spot where the X is located
has two values, the value of the post in dataset R1 and the
value of the post in dataset R2. Since the X is closer to dataset
R2, the value of the R2 post will be more heavily weighted in
the output.
Pixel Alignment and Grid Coordinates
Because GMTED2010 is a multi-resolution dataset, it
is important to know how the variable resolution layers are
related spatially. The layout of the elevation posts for each
resolution (and the corresponding coordinates of the raster
grids) are directly related to the primary source dataset and
to the aggregation approaches used to produce generalized
reduced resolution grids from the source data.
Figure 6. Standard deviation example using
blockstd routine (3 x 3 processing window).
EXPLANATION
No data
8 Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010)
SRTM DTED®2, the primary input source data for
GMTED2010, follows the “pixel center” coordinate referenc-
ing convention in which integer lines of latitude and longitude
that dene the boundary of a 1 x 1-degree tile are located at
the center of a pixel (g. 8). SRTM data have a resolution (or
post spacing) of 0.000277777777 degrees (1 arc-second). By
convention, the SRTM DTED®2 format provides data for a full
1 x 1-degree tile, thereby resulting in a raster grid with dimen-
sions of 3,601 rows by 3,601 columns. In the native SRTM
data format, which follows the DTED® convention, the top
and bottom rows and the left and right columns of a
1 x 1-degree SRTM tile are replicated in the adjacent tiles.
As part of the aggregation processing, the top row and right
column of each input SRTM tile were excluded, resulting
in non-overlapping adjacent generalized tiles that were later
mosaicked into the global GMTED2010 grids.
The exception to the approach of excluding the top row
and right column of each input tile is the standard deviation
product. For this product a different aggregation function
was used than for the other statistical generalization methods,
and it excluded the bottom row and right column of the full
resolution 1-arc-second SRTM input tiles. Thus, a slightly dif-
ferent processing window was used to calculate the standard
deviation product post values compared to the window used
to calculate the other statistical generalizations. By excluding
the bottom row of the full resolution 1-arc-second SRTM input
tile (rather than the top row), the processing window for the
GMTED2010 standard deviation products was shifted north
one row of SRTM input posts from the window that was used
for the minimum, maximum, mean, and median products.
Limited testing has shown that the effect of this slight offset
in processing window alignment is minimal in the 30- and
15-arc-second standard deviation products, where over 91 per-
cent of the grid values exhibit an “error” of less than or equal
to 1 percent. The percent error was calculated by comparing
Figure 7. ArcGIS mosaic – blend method.
Figure 8. “Pixel center” referencing of full resolution 1-arc-
second SRTM data. The diagram shows the 1-arc-second SRTM
pixels in the lower left and upper right portions of a 1 x 1-degree
SRTM tile. Note that the integer lines of latitude and longitude that
define the 1 x 1-degree tile are located at the center of the pixel.
Figure 9. Spatial nesting of GMTED2010 pixels. (A) Each 30-arc-
second pixel covers the area of 900 1-arc-second SRTM pixels
(a 30 x 30-pixel matrix). Each 30-arc-second pixel (B) is spatially
coincident with four 15-arc-second pixels (C), and each 15-arc-
second pixel is spatially coincident with four 7.5-arc-second pixels
(D). Thus, sixteen 7.5-arc-second pixels coincide exactly with
each 30-arc-second pixel.
EXPLANATION
Output Data Products 9
the GMTED2010 standard deviation product grid values with
values derived from a nonshifted processing window. The
discrepancy between grid values in the GMTED2010 standard
deviation products and those that would exist if the process-
ing window had not effectively shifted the input SRTM tile by
one row will be greatest at the 7.5-arc-second resolution and in
low-relief areas. Users are cautioned to consider these issues
when using the standard deviation products.
In the GMTED2010 30-arc-second products, each 30-arc-
second (0.008333333333 degree) pixel covers the exact area
covered by 900 1-arc-second SRTM pixels (a 30 x 30-pixel
matrix). Likewise, each pixel in the GMTED2010 15-arc-
second (0.004166666666 degree) products covers the exact
area of 225 SRTM pixels (a 15 x 15 matrix), and the 7.5-arc-
second (0.002083333333 degree) pixels each cover the exact
area of a 7.5 x 7.5-pixel matrix of full resolution SRTM pixels.
Additionally, the GMTED2010 30-, 15-, and 7.5-arc-second
pixels are nested spatially with a 4-to-1 relationship between
resolution pairs. Thus, each 30-arc-second pixel is spatially
coincident with four 15-arc-second pixels, which are then each
coincident with four 7.5-arc-second pixels. In summary, the
area covered by 900 SRTM pixels (a 30 x 30 matrix) is exactly
spatially coincident with one 30-arc-second pixel, four 15-arc-
second pixels, and sixteen 7.5-arc-second pixels (g. 9).
The spatial relationship of the input SRTM DTED®2
posts and the output GMTED2010 posts is a result of the
aggregation approach used to generalize the SRTM data into
reduced resolution products, resulting in the coordinate refer-
encing of the SRTM data being retained by the GMTED2010
data. The extents of the GMTED2010 raster grids in geo-
graphic coordinates, the resolutions (pixel dimensions), and
the corresponding grid dimensions (rows by columns) are
listed in table 3. Note that because of the pixel center refer-
encing of the input SRTM data (as described above), the full
extent of each GMTED2010 grid as dened by the outside
edges of the pixels differs from an integer value of latitude or
longitude by 0.000138888888 degree (or 1/2 arc-second)
(g. 10). Users of the legacy GTOPO30 product should
note that the coordinate referencing of GMTED2010 and
GTOPO30 are not the same. In GTOPO30, the integer lines of
latitude and longitude fall directly on the edges of a 30-arc-
second pixel. Thus, when overlaying GMTED2010 with
GTOPO30 a slight shift of 1/2 arc-second will be observed
between the edges of corresponding 30-arc-second pixels.
Output Data Products
The new elevation products include the following: mini-
mum, maximum, mean, median, standard deviation, system-
atic subsample, and breakline emphasis. There were 7 prod-
ucts generated at each of the 3 resolutions, for a total of 21
products. All products are in a geographic coordinate system
Table 3. Geographic extents, resolutions, and raster grid dimensions of GMTED2010 product layers.
[std. dev., standard deviation]
Product
Resolution
(decimal de-
grees)
West extent: mini-
mum X-coordi-
nate (longitude)
South extent:
minimum
Y-coordinate
(latitude)
East extent:
maximum
X-coordinate
(longitude)
North extent:
maximum
Y-coordinate
(latitude)
Rows Columns
30 arc-seconds
Minimum
0.0083333333 -180.0001388888
-56.0001388888
179.9998611111
83.9998611111 16,800
43,200
Maximum -56.0001388888 83.9998611111 16,800
Mean * -90.0001388888 * 83.9998611111 * 20,880 *
Median -56.0001388888 83.9998611111 16,800
Std. dev. -56.0001388888 83.9998611111 16,800
Sample * -90.0001388888 * 83.9998611111 * 20,880 *
Breakline -56.0001388888 83.9998611111 16,800
15 arc-seconds
Minimum
0.0041666666 -180.0001388888 -56.0001388888 179.9998611111 83.9998611111 33,600 86,400
Maximum
Mean
Median
Std. dev.
Sample
Breakline
7.5 arc-seconds
Minimum
0.0020833333 -180.0001388888 -56.0001388888 179.9998611111 83.9998611111 67,200 172,800
Maximum
Mean
Median
Std. dev.
Sample
Breakline
* The mean 30-arc-second and sample 30-arc-second layers contain Greenland and Antarctica (the other layers do not contain these landmasses).
10 Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010)
referenced to the WGS 84 horizontal datum, with the hori-
zontal coordinates expressed in decimal degrees. The vertical
units for the elevation values are integer meters, referenced
in most cases to the EGM96 geoid as the vertical datum. For
areas that were derived from
sources other than SRTM
or SPOT 5 Reference3D
data, the vertical datum var-
ies according to the source
data (see table 2b). Data for
Greenland and Antarctica are
referenced vertically to the
EGM96 geoid.
Different products
will be used in a variety of
application situations. For
example, the maximum eleva-
tion product could be used
for the global calculation of
airport runway surface heights
or to determine the height
of vertical obstructions. The
minimum elevation product is
useful for determining stream
channel areas and the water
surface. Comparison of the
minimum and maximum prod-
ucts will provide a measure of
the local relief in a given area.
The standard deviation prod-
uct provides a measure of the
texture, or local variation in elevation, of the landscape sur-
face. The breakline emphasis products will be useful for most
hydrologic applications that involve watershed extraction and
surface streamline routing. The remaining products, speci-
cally the mean and systematic
subsample products, will be
useful for general visualiza-
tion exercises and all-purpose
morphological processing. The
three spatial resolutions allow
the user to choose a level of
detail and corresponding data
volume that are appropriate
for a specic application.
The new generalized
products provide more topo-
graphic detail than the existing
GTOPO30 dataset because
of the introduction of higher
resolution data sources like
SRTM. An area in north-
western Australia was evalu-
ated using the GMTED2010
30-arc-second mean elevation
product and the 30-arc-second
Figure 10. The GMTED2010 layer extents (minimum and maximum latitude and longitude) are a result
of the coordinate system inherited from the 1-arc-second SRTM data.
Figure 11. Comparison of the
existing GTOPO30 and new
GMTED2010 30-arc-second mean
elevation product.
Output Data Products 11
Figure 12. GMTED2010 spatially referenced metadata with void polygons having a dark dense
appearance.
Table 4a. Spatially referenced metadata attribute data
dictionary.
(Created from source elevation data)
Field Description
Source_Org Dataset owner (organization)
Source Type of data ingested
El_Surface Elevation surface type: bare earth or reective
North Upper left Y coordinate
South Lower right Y coordinate
West Upper left X coordinate
East Lower right X coordinate
X_Srce_Res Pixel resolution (X direction)
Y_Srce_Res Pixel resolution (Y direction)
Horz_Unit Horizontal unit
Coord_Sys Spatial reference
Horz_Datum Horizontal datum
Vert_Datum Vertical datum
Vert_Unit Vertical unit
Min_Elev Minimum elevation value
Max_Elev Maximum elevation value
Mean_Elev Mean elevation value
Sdev_Elev Standard deviation elevation value
Prod_Date Metadata creation date
Table 4b. Example values - GMTED2010 spatially referenced
metadata.
(Created from source elevation data)
Field Value
FID 792
Shape Polygon
ID 793
SOURCE_ORG NGA
SOURCE SRTM DTED2 void lled
EL_SURFACE Reective
NORTH -27
SOUTH -28
WEST 121
EAST 122
X_SRCE_RES 1
Y_SRCE_RES 1
HORZ_UNIT Second
COORD_SYS Geographic
HORZ_DATUM WGS 84
VERT_DATUM EGM 96
VERT_UNIT Meter
MIN_ELEV 307
MAX_ELEV 629
MEAN_ELEV 497.64
SDEV_ELEV 45.231
PROD_DATE 31May2008
12 Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010)
GTOPO30 product to contrast the differences in elevation
and topographic detail (g. 11). The difference map in the
upper right corner of gure 11 represents the GMTED2010
30-arc-second mean product minus GTOPO30 and displays
positive differences (GMTED2010 elevation is higher) in red
and negative differences (GMTED2010 elevation is lower) in
blue. In this example, the GMTED2010 30-arc-second mean
product is on average 18 meters lower than GTOPO30. The
new GMTED2010 mean elevation product displays more
pronounced topographic detail in all areas, but especially in
the regions with ats and ridges.
Because the input data used for GMTED2010 come from
multiple sources, spatially referenced metadata have been
generated to accompany the output GMTED2010 elevation
datasets (g. 12). The spatially referenced metadata are con-
tained within a geospatial polygon dataset that contains foot-
prints of each of the source dataset input areas. The attributes
of the source footprint polygons describe the characteristics of
each input dataset used to generate the suite of GMTED2010
products (tables 4a and 4b).
Data Formats
Generalized Elevation Products – Seamless
Global Coverage (GRID)
Global raster data are provided in the ESRI ArcGrid for-
mat. The grid denes geographic space as an array of equally
sized square grid cells (pixels) arranged in rows and columns.
Each grid cell stores a numeric value that represents a geo-
graphic attribute (such as elevation or surface slope) for that
unit of space. Each grid cell is referenced by its x,y coordinate
location. Many commercially available and freeware software
packages can read the ArcGrid format. Detail about le names
and sizes is provided in appendix 1. North of 84 degrees the
void value is assigned to “NoData.”
Generalized Elevation Products – Tile-Based
(GeoTIFF Image Format)
All tiled raster data are provided in the GeoTIFF image
format. GeoTIFF is an open source metadata standard which
allows georeferencing information to be embedded within a
TIFF raster image le. The additional information includes
map projection, coordinate system, ellipsoid, datums, and
everything else necessary to establish the exact spatial refer-
ence for the le. The GeoTIFF format is fully compliant with
TIFF version 6.0, so software that does not read or interpret
the specialized metadata will still be able to open a GeoTIFF
image le. North of 84 degrees the void value is assigned to
-32768.
Spatially Referenced Metadata (ESRI Vector
Shapefile Format)
The spatially referenced metadata are provided in the
ESRI Shapele format. The ESRI Shapele is a vector data
format developed and regulated by ESRI as an open specica-
tion for data interoperability among ESRI and other software
products. A “shapele” commonly refers to a collection of
les with “.shp,” “.shx,” “.dbf,” and other extensions with
a common prex name. The actual vector data are stored in
the le with the “.shp” extension; however, this le alone is
incomplete for distribution because the other supporting les
are required.
Accuracy Assessment
Raster-Based Assessment
A raster-based comparison was conducted between
GMTED2010 and GTOPO30. The calculated minimum, maxi-
mum, mean difference, and standard deviation for the 30-arc-
second systematic subsample and mean products compared to
GTOPO30 are presented in table 5. The mean difference for
both GMTED2010 30-arc-second products indicates about
a 4.3-meter bias compared to GTOPO30. One likely reason
for the positive bias is that the elevation of SRTM areas with
foliage represents vegetation canopy elevations. The stan-
dard deviation of the differences is approximately 91 meters
between GTOPO30 and the GMTED2010 30-arc-second
products, indicating considerable variation in the elevation
differences between the datasets. The error associated with the
elevation differences could also be attributed to the different-
source data used in each elevation model.
Table 5. GMTED2010 global accuracy assessment: raster-
based comparison: GMTED2010 30-arc-second products minus
GTOPO30 (meters).
[GMTED2010, global multi-resolution terrain elevation data 2010;
GTOPO30, Global 30-arc-second elevation dataset]
Minimum Maximum Mean
difference
Standard
deviation
GMTED2010
30-arc-second
systematic
subsample mi-
nus GTOPO30
-4,118 3,314 4.394 91.440
GMTED2010
30-arc-second
mean minus
GTOPO30
-4,130 3,311 4.381 90.124
Accuracy Assessment 13
Absolute Vertical Accuracy Assessment
The absolute vertical accuracy of the GMTED2010 prod-
ucts was measured by spatial comparison with a control point
dataset from the NGA. The control dataset contains nearly
1.6 million coordinate points (latitude/longitude) photogram-
metrically derived from optical stereo imagery. The vertical
accuracy of the control points is better than 10 meters at
90 percent condence, or approximately 6-meter root mean
square error (RMSE). Using the GMTED2010 systematic
subsample 30-arc-second product as a base starting point, all
control points that were located within water or along the land/
water interface and that had an elevation value of zero meters
were removed. There were 64,200 control points with a value
of zero meters from GMTED2010 that were also located
within the land/water interface. Using all non-zero control
points, the standard deviation of the differences was calculated
from the systematic subsample product, and a standard
deviation threshold of 3 was applied to the differences
(table 6). This resulted in a nal set of control points with
outliers removed. The purpose of the threshold is to remove
outliers that are likely present in the reference control point
dataset.
Using the thresholded control points, elevation values
were extracted from the product layers (breakline emphasis,
systematic subsample, mean, and median) at the 30-, 15-,
and 7.5-arc-second resolutions. The Extract Values to Points
tool in ArcGIS was used to extract the elevation value at each
control point location. During the extract process, an interpola-
tion option within the tool was applied. This option allows the
value of the DEM at the control point location to be calcu-
lated from the adjacent pixels with valid values using bilinear
interpolation. The statistical package “R” was then used to
compute the aggregate validation statistics. (“R” is a language
and environment for statistical computing and graphics that
provides a wide variety of statistical and graphical techniques
and is highly extensible.) The rst step was to import the con-
trol point shapele containing the extracted elevation values.
Following the import, the elevation difference between the
GMTED2010 product layers (breakline emphasis, systematic
subsample, mean, and median) and the NGA control points
was computed. The minimum, maximum, mean, and standard
deviation along with the RMSE values were then calculated.
The statistics were aggregated and the overall vertical accu-
racy for each product layer was calculated along with a further
breakdown based on the input source type.
The minimum error, maximum error, mean error, stan-
dard deviation of errors, and RMSE for the GMTED2010
product layers (breakline emphasis, systematic subsample,
mean, and median) derived from a comparison with more
than 1.5 million global control points are presented in
table 7. Because GMTED2010 contains 21 different raster
elevation products at three spatial resolutions, there will not
be one accuracy assessment number associated with the entire
product suite. Instead, four of the individual products (break-
line emphasis, systematic subsample, mean, and median)
will have an accuracy assessment number. The other three
GMTED2010 products (minimum, maximum, and standard
deviation) were not evaluated against the reference control
points because these products were not generated with aggre-
gation methods that select representative elevation values, but
with methods that describe the spread of the elevation values.
The systematic subsample product proved to have the
lowest RMSE at the 30-arc-second, 15-arc-second, and
7.5-arc-second resolutions, probably because the systematic
subsample generalization is selecting actual discrete values
from input source data. Taking the full-resolution SRTM
DTED®2 as an example, 900 1-arc-second elevation values
were used to calculate the value for the single corresponding
elevation at the 30-arc-second resolution. As expected, the
breakline emphasis product had the highest RMSE values at
all three resolutions because the breakline algorithm explic-
itly selects elevation values away from the mean by using the
minimum, maximum, and median values from the input source
data. The mean error in all product layers indicates a small
positive bias with the exception that the breakline emphasis
product is negative at the 30- and 15-arc-second resolutions.
The breakline emphasis method likely gives preference to
lower elevations to enforce stream drainage patterns. The
7.5-arc-second products as a group had the lowest RMSE
values, which is expected because of their resolution, but their
overall accuracy numbers were relatively close to those of the
15-arc-second and two of the 30-arc-second products. The
accuracy results of GTOPO30 when compared to the control
points are listed in table 7. All of the evaluated GMTED2010
product layers surpassed the absolute vertical accuracy of
GTOPO30.
GMTED2010 was constructed from 11 different raster-
based sources. The control points spatially cover all the
continents, with the exception of Antarctica and where the
Australian GEODATA 9-second DEM was used as a void-ll
source. Accuracy assessments for the GMTED2010 product
Table 6. Removal of outliers beyond three standard deviations from the mean difference between NGA
control point dataset and the GMTED2010 systematic subsample product.
Total
non-zero
control point
Standard
deviation of
difference
3 Standard
deviation
threshold
Total
outlier
points
Percent
of outlier
points
Total
final control
points
Systematic subsample 30-arc-
second resolution
1,592,053 60.9365 182.8095 42,353 2.66 1,549,700
14 Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010)
layers (breakline emphasis, systematic subsample, mean, and
median) as a function of the original data source used to pro-
duce GMTED2010 (either to void-ll the 1-arc-second SRTM
data or to provide source data for areas north of 60 degrees
latitude) are presented in tables 8a through 8i. Upon review,
even at their reduced resolution, all of the 7.5-arc-second
products derived from the 1-arc-second SRTM DTED®2 void-
lled data have a vertical accuracy within about ve meters
of the 9.7-meter RMSE specication that was stated for the
1-arc-second SRTM mission (table 8a). At 30, 15, and
7.5 arc-seconds the breakline emphasis products have the
highest RMSE of the four products evaluated, which is the
same pattern recognized in the aggregated global data. In most
cases, the overall bias is positive, reafrming the vegetation
inuence from SRTM. At the 30-arc-second resolution, the
vertical accuracy of the generalized SRTM source areas is
close to that of the global aggregated products (table 7). On
the other hand, at the 7.5- and 15-arc-second resolutions, in
areas where no voids are present, the vertical accuracy of
the GMTED2010 product suite in the SRTM source areas
(table 8a) is much better than the accuracy of the aggregated
global data (table 7).
To further examine the general vertical accuracy pat-
tern from the raster-based sources, the RMSE statistic of the
mean product at 30, 15, and 7.5 arc-seconds is highlighted
in gures 13–15. The global aggregated RMSE for the mean
product at each resolution is also included in the gures for
comparison. Both the RMSE for each source and the number
of control points for each source are displayed. These gures
include GTOPO30 where it was used as source for lling
voids in the 1-arc-second SRTM data. There is a high RMSE
value for the GTOPO30 ll data at all three resolutions. (Note
that GTOPO30 was used only as a last resort source for ll-
ing voids when nothing else was available.) As evident in
the gures, very few control points (1,009 out of 1,549,700)
were located over GTOPO30-lled areas. Although very little
GTOPO30 data were used in the GMTED2010 product suite
(g. 16), the overall product accuracy was affected where
the GTOPO30 data were used. Additionally, areas where
GTOPO30 was used for void-lling coincided with regions
where GTOPO30 was less accurate, such as in South America
and Indonesia. The removal of GTOPO30 as a void-ll source
in any future updates would improve the global aggregate
statistics. At 30 arc-seconds, the RMSE values of only three
sources (CDED3, NED-Alaska, and GTOPO30 ll data) were
above the RMSE for the GMTED2010 mean product (global
aggregate).
These three sources were either created from older carto-
graphic data or were photogrammetrically derived. Likewise,
at 15-arc-second resolution, only NED-Alaska and GTOPO30
Table 7. GMTED2010 absolute accuracy assessment: aggregated global data evaluation: GMTED2010
products and GTOPO30 minus NGA control points (meters).
[GTOPO30, global 30-arc-second elevation dataset; arc-secs, arc-seconds; standard dev, standard deviation; RMSE, root
mean square error]
Breakline
emphasis
Systematic
subsample Mean Median GTOPO30
Control points 1,549,700 1,549,700 1,549,700 1,549,700 1,549,700
30 arc-secs
Minimum -2,874.18 -617.63 -525.58 -2,874.18 -2,551.96
Maximum 690.27 446.03 420.90 433.27 2,944.80
Mean -1.95 3.40 3.91 2.50 5.56
Standard dev 41.25 25.08 26.43 35.32 65.85
RMSE 41.30 25.31 26.72 35.41 66.09
15 arc-secs
Minimum -2,874.18 -2,874.18 -2,874.18 -2,874.18
Maximum 954.39 1,068.27 570.27 458.61
Mean -1.01 1.60 1.89 1.45
Standard dev 31.24 29.12 29.56 29.22
RMSE 31.25 29.17 29.62 29.25
7.5 arc-secs
Minimum -2,874.18 -2,874.18 -2,874.18 -2,874.18
Maximum 632.75 560.27 559.27 563.27
Mean 1.21 0.95 1.02 0.88
Standard dev 29.14 26.78 26.96 26.85
RMSE 29.16 26.80 26.98 26.86
Accuracy Assessment 15
Table 8b. GMTED2010 absolute accuracy assessment by source: CDED1: GMTED2010
products minus NGA control points (meters).
[arc-secs, arc-seconds; standard dev, standard deviation; RMSE, root mean square error]
Breakline
emphasis
Systematic
subsample Mean Median
Control points 13,799 13,799 13,799 13,799
30 arc-secs
Minimum -328.31 -179.85 -241.66 -233.66
Maximum 437.23 176.35 159.46 157.46
Mean -2.62 0.56 0.51 0.06
Standard dev 21.57 16.70 17.87 17.07
RMSE 21.73 16.71 17.88 17.07
15 arc-secs
Minimum -165.63 -142.10 -141.66 -140.10
Maximum 182.54 104.05 100.46 100.05
Mean -1.42 0.39 0.36 0.21
Standard dev 11.09 9.68 10.35 9.88
RMSE 11.18 9.68 10.36 9.89
7.5 arc-secs
Minimum -157.37 -73.37 -70.10 -72.37
Maximum 124.11 106.35 63.30 68.30
Mean -0.17 0.97 0.57 0.55
Standard dev 7.92 6.95 6.67 6.49
RMSE 7.92 7.01 6.69 6.52
Table 8a. GMTED2010 absolute accuracy assessment by source: SRTM DTED®2:
GMTED2010 products minus NGA control points (meters).
[arc-secs, arc-seconds; standard dev, standard deviation; RMSE, root mean square error]
Breakline
emphasis
Systematic
subsample Mean Median
Control points 1,427,975 1,427,975 1,427,975 1,427,975
30 arc-secs
Minimum -595.21 -617.63 -525.58 -508.63
Maximum 690.27 446.03 414.70 433.27
Mean -1.83 3.51 4.01 3.00
Standard dev 33.09 25.06 26.12 24.96
RMSE 33.14 25.30 26.43 25.14
15 arc-secs
Minimum -586.79 -955.58 -741.58 -960.58
Maximum 954.39 1,068.27 570.27 458.61
Mean -0.65 2.06 2.37 1.90
Standard dev 19.09 15.06 15.91 15.25
RMSE 19.10 15.20 16.09 15.37
7.5 arc-secs
Minimum -765.58 -1,225.58 -1,156.58 -1,225.58
Maximum 631.79 560.27 559.27 563.27
Mean 1.65 1.37 1.45 1.30
Standard dev 15.27 9.74 10.26 9.94
RMSE 15.36 9.83 10.36 10.02
16 Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010)
Table 8c. GMTED2010 absolute accuracy assessment by source: CDED3: GMTED2010
products minus NGA control points (meters).
[arc-secs, arc-seconds; standard dev, standard deviation; RMSE, root mean square error]
Breakline
emphasis
Systematic
subsample Mean Median
Control points 10,863 10,863 10,863 10,863
30 arc-secs
Minimum -411.31 -185.15 -402.31 -412.31
Maximum 421.45 181.56 284.16 280.16
Mean -3.87 0.43 0.11 -0.79
Standard dev 34.88 30.11 32.67 31.24
RMSE 35.09 30.11 32.66 31.25
15 arc-secs
Minimum -218.43 -213.42 -267.31 -225.52
Maximum 182.59 169.28 165.28 169.28
Mean -3.36 -0.32 -0.37 -0.51
Standard dev 23.52 22.35 23.22 22.73
RMSE 23.76 22.35 23.22 22.74
7.5 arc-secs
Minimum -835.53 -222.52 -216.52 -217.52
Maximum 632.75 197.94 195.57 196.57
Mean -0.23 0.29 -0.18 -0.25
Standard dev 39.59 21.14 20.80 20.72
RMSE 39.59 21.15 20.80 20.72
Table 8d. GMTED2010 absolute accuracy assessment by source: DTED®1:
GMTED2010 products minus NGA control points (meters).
[arc-secs, arc-seconds; standard dev, standard deviation; RMSE, root mean square error]
Breakline
emphasis
Systematic
subsample Mean Median
Control points 100,371 100,371 100,371 100,371
30 arc-secs
Minimum -425.80 -281.24 -256.34 -281.24
Maximum 433.80 267.35 247.74 250.35
Mean 1.33 1.87 2.17 1.39
Standard dev 27.17 23.34 23.61 22.96
RMSE 27.21 23.41 23.71 23.00
15 arc-secs
Minimum -281.24 -281.24 -281.24 -281.24
Maximum 266.94 266.94 240.94 264.94
Mean -0.07 1.15 1.29 1.11
Standard dev 18.40 17.27 17.55 17.33
RMSE 18.40 17.30 17.60 17.37
7.5 arc-secs
Minimum -281.24 -281.24 -281.24 -281.24
Maximum 372.94 375.94 354.94 372.94
Mean 1.12 0.94 0.94 0.93
Standard dev 15.80 15.53 15.61 15.55
RMSE 15.84 15.55 15.64 15.57
Accuracy Assessment 17
Table 8e. GMTED2010 absolute accuracy assessment by source: GTOPO30 fill data:
GMTED2010 products minus NGA control points (meters).
[arc-secs, arc-seconds; standard dev, standard deviation; RMSE, root mean square error]
Breakline
emphasis
Systematic
subsample Mean Median
Control points 1,009 1,009 1,009 1,009
30 arc-secs
Minimum -575.79 -617.63 -525.58 -508.63
Maximum 463.96 391.45 414.70 416.70
Mean -6.88 -0.10 2.97 1.34
Standard dev 103.64 93.39 97.47 96.23
RMSE 103.82 93.34 97.47 96.20
15 arc-secs
Minimum -586.79 -955.58 -741.58 -960.58
Maximum 438.23 473.61 381.45 458.61
Mean -3.58 0.48 1.17 -0.38
Standard dev 87.07 96.35 89.78 95.67
RMSE 87.10 96.30 89.74 95.63
7.5 arc-secs
Minimum -765.58 -1,225.58 -1,156.58 -1,225.58
Maximum 514.61 489.61 445.61 486.61
Mean 4.30 0.90 0.93 0.91
Standard dev 89.36 102.58 97.49 101.51
RMSE 89.42 102.53 97.44 101.46
Table 8f. GMTED2010 absolute accuracy assessment by source: NED: GMTED2010
products minus NGA control points (meters).
[arc-secs, arc-seconds; standard dev, standard deviation; RMSE, root mean square error]
Breakline
emphasis
Systematic
subsample Mean Median
Control points 286 286 286 286
30 arc-secs
Minimum -31.65 -34.65 -30.65 -31.65
Maximum 16.78 9.46 9.78 10.78
Mean -4.49 -4.31 -4.30 -4.29
Standard dev 7.02 6.77 6.66 6.74
RMSE 8.32 8.02 7.92 7.98
15 arc-secs
Minimum -30.65 -29.62 -30.65 -31.65
Maximum 9.46 8.46 8.46 8.46
Mean -4.10 -4.12 -4.18 -4.15
Standard dev 6.88 6.72 6.71 6.79
RMSE 8.00 7.87 7.90 7.95
7.5 arc-secs
Minimum -37.65 -33.65 -32.65 -33.65
Maximum 12.78 8.46 8.46 8.46
Mean -4.07 -4.17 -4.14 -4.13
Standard dev 7.01 6.82 6.81 6.83
RMSE 8.09 7.99 7.96 7.97
18 Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010)
Table 8g. GMTED2010 absolute accuracy assessment by source: NED-Alaska:
GMTED2010 products minus NGA control points (meters).
[arc-secs, arc-seconds; standard dev, standard deviation; RMSE, root mean square error]
Breakline
emphasis
Systematic
subsample Mean Median
Control points 13,485 13,485 13,485 13,485
30 arc-secs
Minimum -473.03 -182.03 -407.03 -433.31
Maximum 564.00 182.91 420.90 387.90
Mean 8.84 4.75 8.15 6.53
Standard dev 61.13 31.18 54.99 54.03
RMSE 61.76 31.54 55.58 54.42
15 arc-secs
Minimum -260.33 -261.63 -267.31 -269.63
Maximum 248.00 233.90 228.04 232.04
Mean 2.12 4.29 4.50 4.15
Standard dev 30.81 30.77 31.24 30.83
RMSE 30.88 31.06 31.56 31.11
7.5 arc-secs
Minimum -164.63 -164.63 -163.63 -164.63
Maximum 234.10 155.46 155.46 155.46
Mean 2.73 3.21 3.14 3.05
Standard dev 18.77 18.75 18.93 18.69
RMSE 18.97 19.03 19.18 18.94
Table 8h. GMTED2010 absolute accuracy assessment by source: 15-arc-second
SPOT5 Reference3D: GMTED2010 products minus NGA control points (meters).
[arc-secs, arc-seconds; standard dev, standard deviation; RMSE, root mean square error]
Breakline
emphasis
Systematic
subsample Mean Median
Control points 2,574 2,574 2,574 2,574
30 arc-secs
Minimum -419.30 -196.09 -238.09 -238.09
Maximum 255.91 169.76 221.40 229.40
Mean 2.96 1.87 1.75 1.28
Standard dev 31.19 22.32 23.85 23.81
RMSE 31.33 22.39 23.91 23.84
15 arc-secs
Minimum -196.24 -183.77 -183.77 -180.77
Maximum 225.99 234.40 204.40 233.40
Mean 3.86 2.15 2.11 1.98
Standard dev 21.69 17.17 17.41 17.66
RMSE 22.03 17.30 17.53 17.77
7.5 arc-secs
Minimum -156.24 -132.14 -131.14 -133.14
Maximum 210.54 230.40 186.40 228.40
Mean 4.11 1.46 1.33 1.42
Standard dev 18.40 18.31 17.83 18.18
RMSE 18.85 18.36 17.87 18.23
Accuracy Assessment 19
Figure 13. GMTED2010 mean product: 30-arc-second error statistics by source (points in red
are the number of control points used for each source).
7.92
17.88 23.91 23.71
26.43 26.72 32.66
55.58
97.47
0
150,000
300,000
450,000
600,000
750,000
900,000
1,050,000
1,200,000
1,350,000
1,500,000
1,650,000
0
10
20
30
40
50
60
70
80
90
100
110
NATIONAL GEOSPATIAL-INTELLIGENCE
AGENCY CONTROL POINT COUNT BY SOURCE
GMTED2010 versus NGA control points
ROOT MEAN SQUARE ERROR
(RMSE), IN METERS
Figure 14. GMTED2010 mean product: 15-arc-second error statistics by source (points in red
are the number of control points used for each source).
7.95 10.36
16.09 17.53 17.60 23.22
29.62 31.56
89.74
0
150,000
300,000
450,000
600,000
750,000
900,000
1,050,000
1,200,000
1,350,000
1,500,000
1,650,000
0
10
20
30
40
50
60
70
80
90
100
110
GMTED2010 versus NGA control points
NATIONAL GEOSPATIAL-INTELLIGENCE
AGENCY CONTROL POINT COUNT BY SOURCE
ROOT MEAN SQUARE ERROR
(RMSE), IN METERS
Table 8i. GMTED2010 absolute accuracy
assessment by source: Greenland:
GMTED2010 products minus NGA control
points (meters).
[arc-secs, arc-seconds; standard dev, standard devia-
tion; RMSE, root mean square error]
Systematic
subsample
Control points 807
30 arc-secs
Minimum -182.66
Maximum 182.22
Mean 14.13
Standard dev 74.56
RMSE 75.85
20 Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010)
Figure 15. GMTED2010 mean product: 7.5-arc-second error statistics by source (points in
red are the number of control points used for each source).
6.69 7.97 10.36
15.64 17.87 19.18 20.80
26.98
97.44
0
150,000
300,000
450,000
600,000
750,000
900,000
1,050,000
1,200,000
1,350,000
1,500,000
1,650,000
0
10
20
30
40
50
60
70
80
90
100
110
NATIONAL GEOSPATIAL-INTELLIGENCE
AGENCY CONTROL POINT COUNT BY SOURCE
ROOT MEAN SQUARE ERROR
(RMSE), IN METERS
GMTED2010 versus NGA control points
Figure 16. GMTED2010: percent land area by source.
Figure 17. GMTED2010: global aggregated product accuracy.
69.92
13.80 8.71
2.26 2.24 1.79 1.01 0.16 0.09 0.01 0.0004
0
10
20
30
40
50
60
70
80
PERCENT AREA
GMTED2010
Percent land area by source
41.30
25.31
35.41
66.09
31.25 29.17 29.62 29.25
29.16
0
10
20
30
40
50
60
70
Breakline
emphasis
Systematic
subsample
Mean Median GTOPO30
GMTED2010 versus NGA control points
30 arc-seconds
15 arc-seconds
7.5 arc-seconds
ROOT MEAN SQUARE ERROR,
(RMSE), IN METERS
26.80 26.72 26.98 26.86
Summary 21
ll data were above the overall mean product RMSE. The
effect of using GTOPO30 is quite noticeable in the 7.5-arc-
second mean product; if GTOPO30 were removed in these
locations, the overall product RMSE would be closer to
21 meters instead of its current mark of 27 meters. Even accu-
rate data at the 30-arc-second resolution would have a higher
RMSE when represented at the 7.5-arc-second resolution
because the control points are likely to be more distant from
the original source location. Another recognizable pattern is
that the four primary raster-based sources (CDED1, CDED3,
DTED®1, and NED-Alaska) that were used to populate land
areas north of 60 degrees all had vertical accuracy estimates
close to or under the overall mean product RMSE, except for
CDED3 at the 30-arc-second resolution and NED-Alaska at
the 15- and 30-arc-second resolutions.
Finally, sources such as NED and CDED1 that are con-
structed from large-scale cartographic and remotely sensed
imagery inputs have the lowest RMSE values in all three reso-
lutions. NED, a bare-earth DEM source, and CDED1, mostly
a bare-earth DEM, provide a good t relationship with the
absolute control points. Overall, the global aggregated vertical
accuracy of GMTED2010 can be summarized in terms of the
resolution and RMSE of the products (table 7). At 30 arc-sec-
onds, the RMSE range is between 25 and 42 meters; at 15 arc-
seconds, the RMSE range is between 29 and 32 meters; and at
7.5 arc-seconds, the RMSE range is between 26 and 30 meters
(g. 17). These vertical accuracy results are an improvement
over the existing global 30-arc-second digital elevation model
GTOPO30, with a vertical accuracy against the same refer-
ence control points of 66 meters RMSE. If absolute vertical
accuracy is considered as a primary dataset characteristic,
the 15-arc-second resolution may have the most value in the
GMTED2010 product suite. Its vertical accuracy is relatively
close to that of the 7.5-arc-second resolution products, but
the 15-arc-second products have only one quarter of the data
volume. In addition to the improvement in vertical accuracy
over GTOPO30, the overall strengths of GMTED2010 include
more currency and an increased level of detail within the 30-,
15-, and 7.5-arc-second spatial resolutions.
Limitations and Caveats
Depending on the input data source, some artifacts are
apparent in the generalized elevation products. For example,
some areas derived from non-SRTM DTED®1, especially in
Eurasia, exhibit a striping artifact, most likely because of the
production method of the DTED®. The artifact is evident in
the full resolution data, but remains noticeable even in the
generalized elevation product versions. Another pattern seen
in some areas derived from DTED® and CDED is a blocky
appearance, which is a reection of the 1-degree tiling struc-
ture of the full resolution DTED® and CDED. Any area that
was void-lled with GTOPO30 may be problematic because
of the relatively poorer quality of the GTOPO30 source data.
The accompanying spatially referenced metadata can be used
to quickly identify those areas. Fortunately, the land area
void-lled with GTOPO30 is very small and represents only
0.09 percent of the global land surface.
The absolute minimum and maximum elevations for
certain continents may not match published numbers. For
example, in South America and North America the maximum
values are higher than recorded actual land values because
of slight anomalies that were introduced during the lling
of the SRTM data voids with the best available input source
data. The best available ll source data in South America was
GTOPO30, while the 3-arc-second CDED data was the best
for certain areas in North America.
Multiple algorithms and functions within ArcGIS were
used to process and generalize the full resolution source data
into the reduced resolution elevation products. Output data
coordinates of the generalized products appear to vary slightly
between the various ArcGIS algorithms and functions. For
example, the nearest neighbor resampling function was used
to create the systematic subsample product, and its coordinates
are slightly different from those products created using the
aggregate function. All generalized products have the same
number of rows and columns but some truncating or rounding
in the coordinate precision may be observed. Any such differ-
ences in coordinates are only seen in the digits far to the right
of the decimal point in the oating point numeric format of
the coordinates. These small discrepancies in the oating point
format geographic coordinates (decimal degrees) translate to
distances of much less than one meter, which are insignicant
at the spatial resolution of GMTED2010 products.
As with all digital geospatial products, users of
GMTED2010 must be aware of certain characteristics of the
dataset (such as resolution, accuracy, methods of production
and any resulting artifacts) in order to better judge its suitabil-
ity for a specic application. A characteristic of GMTED2010
that renders it unsuitable for one application may have no rel-
evance as a limiting factor for its use in a different application.
Summary
GMTED2010 provides a new level of detail in global
topographic data. Previously, the best available global DEM
was GTOPO30 with a horizontal grid spacing of 30 arc-
seconds. The GMTED2010 product suite contains seven
new raster elevation products for each of the 30-, 15-, and
7.5-arc-second spatial resolutions and incorporates the cur-
rent best available global elevation data. The new elevation
products have been produced using the following aggregation
methods: minimum elevation, maximum elevation, mean
elevation, median elevation, standard deviation of elevation,
systematic subsample, and breakline emphasis. Metadata have
also been produced to identify the source and attributes of all
the input elevation data used to derive the output products.
Many of these products will be suitable for various regional
22 Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010)
and continental applications, such as climate modeling,
continental-scale land cover mapping, extraction of drainage
features for hydrologic modeling, and geometric and radiomet-
ric correction of medium and coarse resolution satellite image
data. The global aggregated vertical accuracy of GMTED2010
can be summarized in terms of the resolution and RMSE of
the products with respect to a global set of control points (esti-
mated global accuracy of 6 m RMSE) provided by NGA. At
30 arc-seconds, the GMTED2010 RMSE range is between
25 and 42 meters; at 15 arc-seconds, the RMSE range is
between 29 and 32 meters; and at 7.5 arc-seconds, the RMSE
range is between 26 and 30 meters. GMTED2010 is a major
improvement in consistency and vertical accuracy over
GTOPO30, which has a 66 m RMSE globally compared to
the same NGA control points. In areas where new sources
of higher resolution data were available, the GMTED2010
products are substantially better than the aggregated global
statistics; however, large areas still exist, particularly above
60 degrees North latitude, that lack good elevation data. As
new data become available, especially in areas that have poor
coverage in the current model, it is hoped that new versions of
GMTED2010 might be generated and thus gradually improve
the global model.
References Cited
Australian Government, 2006, Geoscience Australia, GEO-
DATA 9 Second Digital Elevation Model (DEM-9S)
Version 3 available online at http://www.ga.gov.au/meta/
ANZCW0703011541.html. (Accessed May 6, 2011.)
Australian Government, 2011, Geoscience Australia home
page at http://www.ga.gov.au/. (Accessed May 6, 2011.)
Bamber, J.L., Ekholm, S., and Krabill, W.B., 2001, A new,
high-resolution digital elevation model of Greenland fully
validated with airborne laser altimeter data: Journal of Geo-
physical Research: Solid Earth, v. 106, no. B4,
p. 6,733–6,745.
Bamber, J.L., Gomez-Dans, J.L., and Griggs, J.A., 2008, A
new 1 km digital elevation model of the Antarctic derived
from combined satellite radar and laser data – Part 1: Data
and methods: The Cryosphere Discuss., v. 2, p. 811–841.
Bouillon, A., Bernard, M., Gigord, P., Orsoni, A., Rudowski,
V., and Baudoin, A., 2006, SPOT 5 HRS geometric perfor-
mances—using block adjustment as a key issue to improve
quality of DEM generation: ISPRS Journal of Photogram-
metry and Remote Sensing, v. 60, no. 3, p. 134–146.
Danielson, J.J., and Gesch, D.B., 2008, An enhanced global
elevation model generalized from multiple higher resolution
source datasets: International Archives of the Photogram-
metry, Remote Sensing and Spatial Information Sciences,
v. XXXVII, Part B4, Beijing, p. 1,857–1,864.
ESRI, Mosaic Methods: ArcGIS Resource Center at http://
resources.esri.com/help/9.3/arcgisengine/com_cpp/
gp_toolref/data_management_tools/mosaic_methods.htm.
(Accessed May 6, 2011.)
Farr, T.G., Rosen, P.A., Caro, E., Crippen, R., Duren, R.,
Hensley, S., Kobrick, M., Paller, M., Rodriguez, E., Roth,
L., Seal, D., Shaffer, S., Shimada, J., Umland, J., Werner,
M., Oskin, M., Burbank, D., and Alsdorf, D., 2007, The
Shuttle Radar Topography Mission: Reviews of Geophysics,
v. 45, RG2004, doi:10.1029/2005RG000183.
GeoBase, 2011, Canadian Digital Elevation Data at http://
www.geobase.ca/geobase/en/data/cded/index.html.
(Accessed May 6, 2011.)
Gesch, D.B., 1994, Topographic data requirements for EOS
global change research: U.S. Geological Survey Open-File
Report 94–626, 60 p.
Gesch, D.B., 1999, Chapter 31 – The effects of DEM general-
ization methods on derived hydrologic features, in Lowell,
Kim, and Jaton, Annick, eds., Spatial Accuracy Assess-
ment—Land Information Uncertainty in Natural Resources:
Chelsea, Mich., Ann Arbor Press, p. 255–262.
Gesch, D.B., 2007, The National Elevation Dataset, in Maune,
D., ed., Digital Elevation Model Technologies and Applica-
tions: The DEM Users Manual (2d ed.): Bethesda, Md.,
American Society for Photogrammetry and Remote Sens-
ing, p. 99–118.
Gesch, D.B., and Larson, K.S., 1998, Techniques for devel-
opment of global 1-kilometer digital elevation models, in
Proceedings, Pecora Thirteen, Human Interactions with the
Environment—Perspectives from Space, Sioux Falls, South
Dakota, August 20–22, 1996 (CD-ROM), American Society
of Photogrammetry and Remote Sensing, Bethesda, Md.,
p. 568–572.
Gesch, D.B., Verdin, K.L., and Greenlee, S.K., 1999, New
Land Surface Digital Elevation Model Covers the Earth:
Eos, Transactions, American Geophysical Union, v. 80,
no. 6, p. 69–70.
Government of Canada, Natural Resources Canada, Centre for
Topographic Information, 2007, Canadian Digital Elevation
Data, Level 1 product specications, Edition 3.0: GeoBase,
48 p., available online at http://www.geobase.ca/doc/specs/
pdf/GeoBase_product_specs_CDED1_3_0.pdf.
Grohman, G., Kroenung, G., and Strebeck, J. 2006, Filling
SRTM Voids—The Delta Surface Fill Method: Photogram-
metric Engineering & Remote Sensing, v. 72, no. 3,
p. 213–216.
Institut Geographique National, 2011, Institut Geographique
National home page at http://www.ign.fr/. (Accessed May 6,
2011.)
References Cited 23
Jenson, S.K., and Domingue, J.O., 1988, Extracting Topo-
graphic Structure from Digital Elevation Data for Geo-
graphic Information System Analysis: Photogrammetric
Engineering & Remote Sensing, v. 54, p. 1,593–1,600.
Lehner, B., Verdin, K., and Jarvis, A., 2008, New Global
Hydrography Derived From Spaceborne Elevation Data:
Eos, Transactions, American Geophysical Union, v. 89,
no. 10, p. 93–94.
NASA Jet Propulsion Laboratory, 2009, Shuttle Radar Topog-
raphy Mission home page at http://www2.jpl.nasa.gov/
srtm/. (Accessed May 6, 2011.)
National Geospatial-Intelligence Agency (NGA), 2010. http://
earth-info.nga.mil/GandG/wgs84/gravitymod/ (last refer-
enced Jan. 11, 2011).
National Snow and Ice Data Center, 2011, National Snow and
Ice Data Center home page at http://nsidc.org/. (Accessed
May 6, 2011.)
Slater, J.A., Garvey, G., Johnston, C., Haase, J., Heady, B.,
Kroenung, G., and Little, J., 2006, The SRTM Data “Fin-
ishing” Process and Products: Photogrammetric Engineer-
ing & Remote Sensing, v. 72, no. 3, p. 237–247.
Spot Image, 2008, Astrium GEO-Information Services home
page at http://www.spot.com/?countryCode=US&language
Code=. (Accessed May 6, 2011.)
Spot Image, 2005, Reference3D product description, Ver-
sion 5.2: Spot Image, 17 p., available online at http://
www.spot.com/automne_modules_les/standard/public/
p453_e66fdc8d9baeb629a19beb53be67339dReference3D-
Product_descriptionv5-2.pdf.
U.S. Geological Survey, 2006, National Elevation Dataset
home page at http://ned.usgs.gov/. (Accessed May 6, 2011.)
U.S. Geological Survey, 2011, U.S. Geological Survey home
page at http://www.usgs.gov/. (Accessed May 6, 2011.)
Verdin, K.L., and Greenlee, S.K., 1996, Development of
continental scale digital elevation models and extraction of
hydrographic features, in Proceedings, Third International
Conference/Workshop on Integrating GIS and Environ-
mental Modeling: Santa Fe, N. Mex., January 21–26, 1996,
National Center for Geographic Information and Analysis,
Santa Barbara, Calif.
Zwally, H.J., Schutz, B.E., Bentley, C.R., Bufton, J., Her-
ring, T., Minster, J.F., Spinhirne, J., and Thomas, R., 2007,
GLAS/ ICESat L2 Antarctic and Greenland Ice Sheet Altim-
etry Data V428, September 25, 2003–November 27, 2006,
National Snow and Ice Data Center, Digital Media, Boulder,
Colo.
Appendix
26 Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010)
Appendix 1. Filenames and sizes of GMTED2010 products (global grids).
[GB, gigabytes; Std. deviation, standard deviation]
Resolution Grid name Product Rows x columns Data volume
(GB)
30-arc-seconds
(0.0083333333 deci-
mal degrees)
mi30_grd Minimum 16,800 x 43,200 0.335
mx30_grd Maximum 16,800 x 43,200 .347
mn30_grd Mean 20,880 x 43,200 .417
md30_grd Median 16,800 x 43,200 .340
sd30_grd Std. Deviation 16,800 x 43,200 .201
ds30_grd Sample 20,880 x 43,200 .421
be30_grd Breakline 16,800 x 43,200 .343
15-arc-seconds
(0.0041666666 deci-
mal degrees)
mi15_grd Minimum
33,600 x 86,400
1.22
mx15_grd Maximum 1.25
mn15_grd Mean 1.24
md15_grd Median 1.24
sd15_grd Std. deviation .712
ds15_grd Sample 1.25
be15_grd Breakline 1.25
7.5-arc-seconds
(0.0020833333 deci-
mal degrees)
mi75_grd Minimum
67,200 x 172,800
4.53
mx75_grd Maximum 4.60
mn75_grd Mean 4.57
md75_grd Median 4.57
sd75_grd Std. Deviation 2.48
ds75_grd Sample 4.60
be75_grd Breakline 4.59
Publishing support provided by:
Rolla Publishing Service Center
For more information concerning this publication, contact:
U.S. Geological Survey Earth Resources Observation and Science
(EROS) Center
47914 252nd Street
Sioux Falls, South Dakota 57198
(605) 594-6151
gmted2010@usgs.gov
Or visit the EROS Center Web site at:
http://eros.usgs.gov/
Danielson and Gesch—Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010) —Open-File Report 2011–1073
