Accuracy assessment and bias correction for open sources Digital elevation model (DEM): a case study in and around Diredawa city, Ethiopia
No Thumbnail Available
Date
2022-02
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Digital elevation model (DEM) is a digital and mathematical representation of the real-world topographic surface by using elevation data. The global open-access DEM data contain errors and are not suitable for engineering and other high accuracy requiring applications. The main aims of this study are vertical accuracy assessment and bias correction for open-source global DEM data. Conventional global DEM data; SRTM DEM 1-arc second and ASTER-GDEM 1-arc second, and new global DEM data ALOS-PALSAR both 12.5 m and 30 m spatial resolution DEM data were applied for this study. Ground control point GCP, RTK-GPS data, and photogrammetric DEM reference data were used to evaluate the accuracy of the selected open-source global DEM data. Accuracy assessment was applied in different approaches, one is a point-based approach using statistical measurement. The computed Root mean square (RMSE) for SRTM is 3.68 m, 4.24m, and 3.41m in GCP, RTK-GPS, and Photogrammetric DEM reference data, for ASTER-GDEM the RMSE value is 8.87 m, 7.08 m, and 7.05 m, for ALOS-PALSAR12.5 m it is 4.58 m, 4.46 m and 4.58 m and for ALOS- PALSAR30 m the value is 5.79 m, 5.10 m and 5.34 m, respectively. The accuracy of SRTM showed better in three reference data and ALOS-PALSAR12.5 m is followed the SRTM in the accuracy. Whereas the accuracy of ASTER-GDEM showed less accuracy among the selected global DEM data. The second approach is based on land use land cover data using RTK-GPS reference data; bare land, road line, drainage pattern, rock area, and tree areas point data are collected using TRIMBLE R8 GNSS/R6/5800 instrument, the accuracy of rock area showed less accuracy in all DEM data except SRTM DEM and the accuracy of the bare land area showed better in all DEM averagely. And the third approach of vertical accuracy assessment is terrain classification; the terrain of this study area is segmented into three classes (flat area, moderate slope, and steep slope), the accuracy of the flat area showed high and the accuracy of the steep slope is less for all DEM data. Finally, bias correction was applied using linear transformation parameters to reduce the residual of the global DEM data, the accuracy of ASTER-GDEM was improved by 22.9 % after the bias correction, while the accuracy of SRTM is improved by 6.7 % and the accuracy of PALSAR12.5m and PALSAR30 m was improved by 5.1 % and 2.2 % after the bias correction.
Description
Keywords
Digital Elevation Model, Linear transformation, Reference Data, Vertical accuracy