Evaluation of the atmospheric correction procedure for the APEX level 2/3 processor

摘要

The Airborne Prism Experiment (APEX) is a hyperspectral instrument built in a Swiss - Belgian collaboration within the ESA-PRODEX program [1]. It aims at highest possible accuracy of its delivered surface reflectance image data products.\udThe atmospheric correction of hyperspectral imagery is a critical element of a complete processing chain towards\udunbiased reflectance and for the creation of higher level products. As the first data of APEX is expected to become\udavailable in 2009, an appropriate processing chain for higher level processing needs to be defined and evaluated.\udStandard products have been identified in all application fields of hyperspectral imaging, i.e., geology, vegetation,\udcryosphere, limnology and atmosphere. They are being implemented at the APEX science center [2]. The according\udprocessing procedures rely on data of well-defined processing states which range from calibrated at-sensor radiance to(bihemispherical) spectral albedo.\udIn this paper, the atmospheric processing which is implemented as part of the automated data processing chain for level 2 in the APEX processing and archiving facility (PAF) [3] is evaluated together with the ATCOR-4 atmospheric correction program [4],[5]. The evaluation is done regarding flexibility, reflectance output accuracy and processing efficiency. Two test data sets are taken for this purpose: a well-documented set of HYMAP data [6] and a high resolution HYSPEX data set [7]. Both data sets exhibit areas of overlap, which are taken for self-contained analysis of the atmospheric correction procedure. The accuracy tests include plausibility checks on selected regions of interest including a variety of known surfaces in the imagery. As some of the observed effects are related to BRDF differences, the results also give an indication for the inaccuracy related to these reflectance anisotropies. Speed measurements of the processing are then\udcompared to the demand for operational processing of series of data acquisition. Further comparison information is\uddrawn from the by-products of atmospheric correction such as water vapor distribution maps.\udThe study shows performance and limitations of atmospheric correction using the state-of-the-art technology, which are\udmainly found in the field of BRDF effects. This points towards improvements to be implemented in course of the further development of the higher level processing chain for the APEX sensor.