PRISM/DSM Mosaic Generation and Evaluation

Abstract

The Advanced Land Observing Satellite (ALOS), which was launched in January 2006, ended its successful mission life in May 2011. The satellite had observed global land areas from space for more than five years and left us with a huge amount of data. The Panchromatic Remote-sensing Instrument for Stereo Mapping (PRISM), one of ALOS’s onboard sensors, was designed to generate worldwide topographic data with its stereoscopic, hig-resolution observation. The sensor consists of three independent panchromatic radiometers for forward, nadir, and backward viewing to produce a triplet stereoscopic image along its track. Each radiometer has 2.5 m ground resolution in its 35 km swath. We developed software to generate Digital Surface Model (DSM) data from the stereoscopic images for the Earth Observation Research Center/Japan Aerospace Exploration Agency (EORC/JAXA). The software fully supports the rigorous sensor models that are being applied to standard products. It generates DSMs in 10 m grid spacing using a unique image-matching algorithm developed exclusively for PRISM and then archives them scene by scene. We have processed more than 6000 DSM scenes of Japan and some specific sites that have their validations and research applications. These DSM scenes must be combined and interpolated onto the familiar geographic latitude/longitude grid for the final standard DSM dataset. This paper reports on the status of the archive of DSM scenes on all Japan land areas and the mosaicking algorithm, which combines the data into 1°×1° tiles, along with its processing results. The absolute and relative accuracies of mosaicked DSM tiles are validated with ground control points (GCPs) and a variety of height reference data, including DSMs produced by airborne light detection and ranging (LiDAR) instruments. The absolute height accuracy was confirmed at 2.9 m as the root mean square error using more than 3000 GCPs on the mosaicked DSM tiles.