ASTER GDEM quality analysis
ASTER topography (GDEM V1)
The publication of the ASTER GDEM V1 (30m) is a great step towards a worldwide high resolution elevation model. We have done some test in the Southern Alps around Trento which is among the most complex terrains in Europe. The scope was to calculate a difference map to the local high resolution DEM.
- Mosaicking and reprojection to UTM32/WGS84 of ASTER GDEM tiles of the area with GDAL
- Import of the ASTER GDEM map into GRASS GIS
- Creation of a difference map to the provincial DEM (based on 1m Lidar DEM)
- Histogram and univariate statistics.
(click for higher resolution or download slides as PDF)
The hydrological pattern appears to coincide with the provincial rivers map. It can be seen that lakes weren't masked during the ASTER DEM preparation, howver, these could be identified for many areas in the world with OpenStreetMap layers.
Despite spikes (including unexpected craters), the overall quality appears to be acceptable for this first version of ASTER GDEM - the standard deviation is 18m for the test area with outliers predominantely found in the areas of complex terrain rather than in the valley floors. A shift of -2.3m is also observed (reprojection artifact?). It will be worth to check if SRTM data (90m) could help to identify and remove the spikes from the DEM.
ASTER topography (GDEM V2)
Improved ASTER GDEM 2 from 2011:
The ASTER GDEM covers land surfaces between 83°N and 83°S and is comprised of 22,702 tiles. Tiles that contain at least 0.01% land area are included. The ASTER GDEM is distributed as Geographic Tagged Image File Format (GeoTIFF) files with geographic coordinates (latitude, longitude). The data are posted on a 1 arc-second (approximately 30–m at the equator) grid and referenced to the 1984 World Geodetic System (WGS84)/ 1996 Earth Gravitational Model (EGM96) geoid.
Notes: ASTER GDEM can be rather well filtered and smoothed with the Sun's denoising algorithm (using GDAL and free / open source program <mdenoise> or simply GRASS add-on r.denoise. Experiments showed that the best smoothing of ASTER GDEM 2 is reached with such parameters of 'mdenoise' with threshold = 0.8 and iterations = 10-20. Also filtering with r.neighbors by "average" method and window size >=5 is quite useful to remove some noise from DEM.
For hydrological modelling, ASTER GDEM can also be improved in GRASS GIS 7 with r.hydrodem