Assessment of Accuracy Fluctuations in Multi-Resolution Icesat-2 Terrain and Surface Heights Over Time

Abstract

The Ice Cloud and Land Elevation Satellite-2 (ICESat-2) mission, spanning more than five years, has collected extensive three-dimensional Earth observation data, equipping the scientific community with innovative tools for understanding environmental changes on a global scale. The ATL08, a key product of the ICESat-2 mission, offers vital global land and vegetation height data for carbon budget and cycle modeling. The consistent measurement accuracy of the ATL08 product over time is essential for its applications in time series analyses. However, previous studies have largely ignored fluctuations in the temporal accuracy of ATL08. This study bridges the knowledge gap by analyzing 59 months of the latest release of ATL08 version 006, from October 2018 to August 2023, to assess the accuracy of terrain and surface height retrieval across six major land cover types in Finland. It explores the fluctuations in ATL08 retrieval accuracy over time and employs a random forest (RF) model to evaluate the relative importance of error factors on accuracy and their interaction. Moreover, the study assesses accuracy at two spatial resolutions, i.e., 100 m×11 m and 20 m×11 m, to evaluate the capability of ATL08 for high-resolution retrieval. Across the entire dataset, terrain elevation retrieval exhibits high accuracy, showing significant improvement over the previous ATL08 version. For the 100m segment, the accuracy metrics are a bias of 0.04 m, a MAE of 0.44 m, and a RMSE of 0.66 m. In contrast, the 20m segment reports a bias of 0.10 m, a MAE of 0.35 m, and a RMSE of 0.49 m. However, surface height retrieval accuracy is considerably lower. In the 100m segment, the bias is -0.59 m, MAE is 3.06 m, RMSE is 4.52 m, with a bias% of -3.45%, a MAE% of 21.26%, and a RMSE% of 31.40%. For the 20m segment, the bias stands at -0.72 m, a MAE of 3.51 m, a RMSE of 5.23 m, a bias% of -5.81%, a MAE% of 28.52%, and a RMSE% of 42.47%. The results indicate that improving segment resolution enhances terrain accuracy, although this improvement coincides with a reduction in surface height accuracy. Through the analysis of the relative importance of error factors, surface coverage and beam type are identified as crucial factors affecting terrain retrieval accuracy, exhibiting variable temporal effects. Seasonal changes, especially the presence of snow, affect the accuracy of terrain retrieval, with accuracy being notably lower around March each year. The study confirms the critical impact of surface height on its retrieval accuracy and advises against using ATL08 for low target surface height retrieval, pointing out challenges in steep terrains. Nevertheless, it affirms the applicability of ATL08 for canopy height estimation in boreal forests, predominantly composed of coniferous species, underscoring its potential for extensive spatial and temporal research. This contributes significantly to bridging data gaps between accurate estimates and large area coverage in global-scale carbon budget and cycle studies.

Keywords: ICESat-2, ATL08, time series, Accuracy fluctuations, terrain, Surface height