VALIDATION OF THE ASTER GLOBAL DIGITAL ELEVATION MODEL 
VERSION 2 OVER THE CONTERMINOUS UNITED STATES 
D. Gesch * *, M. Oimoen* "5, Z. Zhang ^ ", D. Meyer, J. Danielson ? 
* U.S. Geological Survey, Earth Resources Observation and Science Center, Sioux Falls, South Dakota, USA 57198 — 
gesch@usgs.gov, dmeyer@usgs.gov, daniels@usgs.gov 
® SGT, Inc., contractor to the USGS Earth Resources Observation and Science Center, Sioux Falls, South Dakota, USA 57198 - 
oimoen@usgs,gov 
* ERT, Inc., contractor to the USGS Earth Resources Observation and Science Center, Sioux Falls, South Dakota, USA 57198 - 
zhang@usgs,gov 
Commission IV, WG IV/6 
KEY WORDS: Accuracy, DEM/DTM, Comparison, Geodesy, Global-Environmental-Databases, Land Cover, Mapping, Satellite 
ABSTRACT: 
The ASTER Global Digital Elevation Model Version 2 (GDEM v2) was evaluated over the conterminous United States in a manner 
similar to the validation conducted for the original GDEM Version 1 (v1) in 2009. The absolute vertical accuracy of GDEM v2 was 
calculated by comparison with more than 18,000 independent reference geodetic ground control points from the National Geodetic 
Survey. The root mean square error (RMSE) measured for GDEM v2 is 8.68 meters. This compares with the RMSE of 9.34 meters 
for GDEM vl. Another important descriptor of vertical accuracy is the mean error, or bias, which indicates if a DEM has an overall 
vertical offset from true ground level. The GDEM v2 mean error of —0.20 meters is a significant improvement over the GDEM v1 
mean error of —3.69 meters. The absolute vertical accuracy assessment results, both mean error and RMSE, were segmented by land 
cover to examine the effects of cover types on measured errors. The GDEM v2 mean errors by land cover class verify that the 
presence of aboveground features (tree canopies and built structures) cause a positive elevation bias, as would be expected for an 
imaging system like ASTER. In open ground classes (little or no vegetation with significant aboveground height), GDEM v2 
exhibits a negative bias on the order of 1 meter. GDEM v2 was also evaluated by differencing with the Shuttle Radar Topography 
Mission (SRTM) dataset. In many forested areas, GDEM v2 has elevations that are higher in the canopy than SRTM. 
1. INTRODUCTION extent of the conterminous United States (CONUS) in a manner 
similar to the validation conducted for the original GDEM v1 in 
The initial version of the ASTER Global Digital Elevation 2009. The primary goal of the CONUS validation work 
Model (GDEM v1) (Abrams et al., 2010), a joint project of the reported here was to fully characterize the vertical accuracy of 
U.S. National Aeronautics and Space Administration (NASA) GDEM v2. 
and the Ministry of Economy, Trade, and Industry (METI) of 
Japan, was released in June 2009. The user community widely 
embraced the availability of GDEM vl even though NASA and 
METI acknowledged it to be a "research grade" dataset that 
contains anomalies and artifacts that may limit its usefulness for 2.1 Validation Approach 
some applications. Numerous validation efforts conducted on 
GDEM v1 concluded that in most cases the dataset met its All of the 934 Ixl-degree tiles of GDEM v2 data covering 
stated accuracy goal (+20 meters at 95% confidence) but that CONUS were included in the validation effort. Absolute 
some characteristics of the dataset affect how the terrain is vertical accuracy of GDEM v2 was calculated by comparison 
represented and how the DEM performs in applications with independent reference geodetic ground control points. 
2. METHODS AND DATA 
  
(ASTER GDEM Validation Team, 2009; Slater et al., 2011; 
Hvidegaard et al., 2012; Miliaresis and Paraschou, 2011; Wang 
et al., 2012). 
To address limitations of GDEM v1, NASA and METI jointly 
developed GDEM Version 2 (v2) (Tachikawa et al., 2011) and 
released it to the user community in October 2011. The 
improvements in the processing for GDEM v2 include an 
additional 260,000 individual ASTER scenes to improve 
coverage, a smaller correlation window to improve spatial 
resolution, and better water masking (ASTER GDEM 
Validation Team, 2011). GDEM v2 was evaluated over the full 
  
* Corresponding author. 
Work performed under USGS contract G10PC00044. 
GDEM v2 was also evaluated by pixel-to-pixel differencing 
with other l-arc-second (30-meter) DEMs that have complete 
coverage over CONUS, namely the National Elevation Dataset 
(NED) (Gesch, 2007) and the Shuttle Radar Topography 
Mission (SRTM) dataset (Farr et al, 2007). Accuracy 
assessment results were segmented by land cover classes to look 
for relationships between vertical accuracy and cover type. One 
characteristic of GDEM v2, specifically the number of ASTER 
scenes (stereo pairs) used to derive an elevation for a pixel, was 
examined to see how it affected vertical accuracy. 
265