Error Characteristics of Pan-Arctic Digital Elevation Models and Elevation Derivatives in Northern Sweden

• Published in: Remote Sensing
• Main Authors: Martin Karlson, David Bastviken, Heather Reese
• Format: Article in Journal/Newspaper
• Language: English
• Published: MDPI AG 2021
• Subjects: digital elevation model; accuracy assessment; elevation derivatives; terrain; Arctic region; land cover; Science; Q; Arctic; Kalix; Northern Sweden
• Online Access: https://doi.org/10.3390/rs13224653; https://doaj.org/article/730eefba70ed45389b0cd4c836f0a737

Many biochemical processes and dynamics are strongly controlled by terrain topography, making digital elevation models (DEM) a fundamental dataset for a range of applications. This study investigates the quality of four pan-Arctic DEMs (Arctic DEM, ASTER DEM, ALOS DEM and Copernicus DEM) within the Kalix River watershed in northern Sweden, with the aim of informing users about the quality when comparing these DEMs. The quality assessment focuses on both the vertical accuracy of the DEMs and their abilities to model two fundamental elevation derivatives, including topographic wetness index (TWI) and landform classification. Our results show that the vertical accuracy is relatively high for Arctic DEM, ALOS and Copernicus and in our study area was slightly better than those reported in official validation results. Vertical errors are mainly caused by tree cover characteristics and terrain slope. On the other hand, the high vertical accuracy does not translate directly into high quality elevation derivatives, such as TWI and landform classes, as shown by the large errors in TWI and landform classification for all four candidate DEMs. Copernicus produced elevation derivatives with results most similar to those from the reference DEM, but the errors are still relatively high, with large underestimation of TWI in land cover classes with a high likelihood of being wet. Overall, the Copernicus DEM produced the most accurate elevation derivatives, followed by slightly lower accuracies from Arctic DEM and ALOS, and the least accurate being ASTER.