Validation of Airborne Hyperspectral Imagery from Laboratory Panel Characterization to Image Quality Assessment: Implications for an Arctic Peatland Surrogate Simulation Site
Calibration/validation (cal/val) practices applied to airborne hyperspectral imagery of Arctic regions were developed and assessed as an integrated up-scaling methodology that considers: (i) calibration of a laboratory reflectance reference panel; (ii) cross-calibration of multiple field panels; (ii...
Published in: | Canadian Journal of Remote Sensing |
---|---|
Main Authors: | , , , |
Format: | Article in Journal/Newspaper |
Language: | English French |
Published: |
Taylor & Francis Group
2019
|
Subjects: | |
Online Access: | https://doi.org/10.1080/07038992.2019.1650334 https://doaj.org/article/54c4f467e79f4f6bb6cd48b7286a3af4 |
Summary: | Calibration/validation (cal/val) practices applied to airborne hyperspectral imagery of Arctic regions were developed and assessed as an integrated up-scaling methodology that considers: (i) calibration of a laboratory reflectance reference panel; (ii) cross-calibration of multiple field panels; (iii) quality assurance checks of field spectroscopy data; and, (iv) comparison of results with airborne hyperspectral imagery. Overall errors of up to 27% were reduced to <4% using these methods. Calibration results of the laboratory panel provided an improvement of 1% in the visible, near and lower shortwave infrared regions with respect to best estimates achievable using manufacturer supplied calibration data. This improvement was transferred to field panels using an in-house cross-calibration approach that also allowed panels to be assessed for degradation that occurs during field deployment. Comparison of the field spectroscopy results of four cal/val targets with hyperspectral imagery following atmospheric correction identified discrepancies from 1% to 4% (absolute) between 450 nm and 1050 nm, with errors as high as 27% at lower wavelengths. Application of scene-based refinements using two cal/val targets reduced this error across the entire spectral range (<4%) with the most significant improvements below 500 nm. These methods also have important implications to satellite image analysis, especially in Arctic and northern regions. |
---|