Automated Processing of MABEL and mATLAS Photon Counting Lidar Data: Initial Steps Toward Global-Scale Forest Measurements From ICESat-2

Wednesday, 17 December 2014
Ryan Sheridan, Texas A & M University, College Station, TX, United States, Sorin C Popescu, Texas A&M University, Department of Ecosystem Science and Management, College Station, TX, United States and Ross Nelson, NASA Goddard Space Flight Center, Greenbelt, MD, United States
ICESat-2 is NASA’s next generation spaceborne laser altimeter, which aims to follow in the footsteps of ICESat. A major difference between this and the previous ICESat mission is the utilization of a photon counting lidar. Such systems are relatively new, and their ability to produce vegetation measurements remains largely unexplored. This study focuses on development and testing of an automated processing algorithm for high altitude photon counting lidar data, utilizing Multiple Altimeter Beam Experimental Lidar (MABEL) and mATLAS data. mATLAS data are simulated data generated utilizing previously collected MABEL data. Conditions present during MABEL data collection are typically static while simulating mATLAS data enables varied noise levels, allowing us to test algorithm robustness. Specific goals of this algorithm include: (1) reduction of noise within the data; (2) delineation of top of canopy (TOC) and ground surfaces; (3) classification of photon returns (e.g. noise, TOC, ground, within canopy); and (4) calculation of above ground level (AGL) heights for signal photons not classified as ground. Algorithm accuracy assessment is performed using spatially coincident MABEL and Goddard’s Lidar, Hyperspectral, and Thermal Airborne Imager (G-LiHT) data. Our algorithm processes data utilizing an overlapping moving window approach, ancillary datasets, median filters, pseudo-waveforms, and spline fitting techniques. Early performance on MABEL data shows lower overall RMSE for ground than TOC (3.26 and 9.58, respectively), and identified high TOC commission error rates in open areas (i.e. areas with little or no vegetation). These TOC commission errors were responsible for increased ground omission errors. After adjusting algorithm parameters, visual assessments of subsequent tests suggest reductions in open area TOC commission and ground omission rates. Overall, ground identification performed well in areas with dense canopy cover. Similar results were observed when processing mATLAS data. A notable exception is data exhibiting weak beam strength conditions, causing signal to become more difficult to identify. Future work will focus on completion of a formal accuracy assessment for mATLAS processing results, and fine-tuning algorithm parameters to deal with varied topographic relief.