Dimension reduction of multi-temporal elevation data

Abstract

Near-global digital elevation models (DEMs) with a horizontal resolution 1 arc sec, from the Shuttle Radar Topography Mission (SRTM DEM), the Advanced Space borne Thermal Emission and Reflection Radiometer (ASTER DEM) sensor, and the Panchromatic Remote-sensing Instrument for Stereo Mapping (PRISM) on board the Advanced Land Observing Satellite (ALOS DEM) are available. These three elevation products present differences in the elevation conceptual framework resulting from noise and error effects as well as to changes in landcover and geomorphology of earth surface. The aim of this research effort is to highlight the differences among these three DEMS. Towards this end, a method is presented that is capable of decomposing the multi-dimensional elevation data formed by these 3 DEMs to two components: (a) the first one (adjusted DEMs) includes the mostly similar/common information among the initial DEMs, while (b) the second one (residual DEMs) maximizes the differences among the data. Adjusted DEMs might be used as first-degree “optimal” terrain representations while the residual DEMs are suitable to outline and map regions of high- and low-elevation variability among the three DEMs under comparison, due to (a) errors, (b) differences in the elevation conceptual framework among the three DEMs, and (c) landcover change over time.