|
Pierre, I am inclined to agree with Jonathon Ferranti's analysis. ASTERGDEM data is overrated and its accuracy exaggerated. Another major problem IMV is that the user cannot have confidence in the elevation data for any given point. What most people do not appear to appreciate is that ASTERGDEM is a realative DEM. In this sense it is not much different from the relative DEMs that can be obtained from Land Processes Distributed Active Archive Center and others. ASTERGDEM was produced using software developed by the Sensor Information Laboratory Corporation (SILC) of Japan. The software does not allow XYZ ground control points to be used to improve the positional accuracy of the elevation data extracted by auto-stereocorrelation of nadir and along-track oblique view data collected by ASTER. The geo-postional accuracy of the XYZ coordinates that make up the ASTERGDEM dataset is purely a function of knowledge about the velocity, track, position and orientation of the ASTER sensor and the shape of the geoid below, when data collection occured. It is an improvement on the relative DEM data available previously because it combines elevation data extracted from multiple scenes. This helps suppress spurious elevation values. However, a lot of these spurious elevation values are a result of limitations of the auto-stereocorrelation, which requires that pixel values around any individual pixel in the nadir and oblique view data exhibit sufficient variation in value for coresponding points within each view to be identified accurately. This commonly fails over uniform and high albedo regions, like sand desert snow and ice. The improvement ASTERGDEM seems to exhibit over relative DEMs available previously, is better XY postional accuracy, and to a lesser degree, better postional accuracy in Z. The former has probably been achieved by making use of confidential information about the satelite postion and motion, or perhaps by better modelling of the information on these in the public domain. The latter seems to have been achieved by merging of data from multiple DEM extractions for the same region, which has the effect of suppressing spurious elevations. However, if stereocorrelation fails because of a no-match being obtained in a nadir and oblique view pair, it is likely to fail in others of the same area, because the physical characteristics of the earth surface commonly remain unchanged.
Dream no small dreams for they have no power to change the minds of men. - After Johann Wolfgang von Goethe The Fermi paradox restated: "The size of TSMC's wafers, improving 45nm yields and the length of its development period suggests that many examples of functioning high-binning Fermi GPGPUs ought now to exist in the Universe. However, this hypothesis seems inconsistent with the lack of observational evidence to support it." |