Extracting Satellite Laser Altimetry Footprints With the Required Accuracy by Random Forest

Abstract

Due to its high elevation accuracy and wide coverage, satellite laser altimetry plays an important role in many scientific fields, such as polar ice sheet monitoring, vegetation canopy height measurement, and topography mapping. However, the elevation accuracy of satellite laser altimetry data is affected by many factors, such as the atmosphere, instrument noise, terrain fluctuation, etc., which leads to an uncertain accuracy. In this letter, to solve this problem, we propose a method based on random forest to extract satellite laser altimetry footprints that meet the elevation accuracy requirements of certain applications in complex terrain. Using ICESat, we take the elevation control point accuracy requirement for 1:10 000 mapping as an example to verify the proposed method. Experimental results show that the elevation root mean square errors (RMSEs) of the selected high-quality footprints are 0.41, 0.70, and 0.87 m in flat land, hills land, and mountainous areas, respectively, which meets the requirements of 1:10 000 topography mapping. The percentage of extracted footprints that meet the elevation accuracy requirement from the three terrains are all higher than 90%.