Ray-matching과 의사불변검보정 사이트를 활용한 절대복사보정의 정확도 평가

Abstract

Absolute radiometric calibration is a fundamental process aimed at converting satellite digital numbers (DN) into physically meaningful radiance or reflectance values. This process is vital for translating satellite observations into data that represent actual radiation energy and spatial characteristics, enabling the monitoring of climate and environmental conditions. Therefore, the calculation of radiometric calibration coefficients is crucial for extracting meaningful information from satellite data post-launch. Traditionally, there are two methods for computing radiometric calibration coefficients: On-board calibrators (OBCs) and vicarious calibration methods. OBCs are typically employed by government-operated satellite systems, but they have limitations when applied to commercial and small satellites due to cost considerations. As a result, vicarious calibration methods, relying on ground-based calibration, are often used. However, ground calibration campaigns are constrained by their fixed locations, resulting in spatial limitations. Therefore, in this study, the absolute radiometric calibration accuracy was evaluated based on other satellite images by utilizing the Spectral Band Adjustment Factor (SBAF) of Ray-matching, which compensates for the spectral difference between two different sensors. In this study, desert and dark targets were selected to calculate the DN to Top-Of-Atmosphere reflectance (TOA reflectance) radiometric calibration coefficients considering a wide range of reflectance values. Landsat-8 was selected as the optical satellite to calculate the radiometric calibration coefficients, and it was performed based on Sentinel-2A. The radiometric calibration coefficient calculation was categorized into two cases: Case 1, which exclusively utilized desert pixels, and Case 2, which incorporated both desert and dark target pixels. Pseudo TOA reflectance was derived through SBAF application, and its applicability was assessed. Coefficients were calculated through linear regression with Landsat-8 DN. The accuracy evaluation of this study indicated that both Case 1 and Case 2 produced valid results, with accuracies within 2% for the blue, green, red, and narrow NIR channels. Notably, in the NIR channel, where there was a substantial bandwidth difference, the TOA reflectance error was reduced by approximately 2% when both desert and dark targets were considered, compared to the case where only desert sites were applied. This method for radiometric calibration coefficient calculation, as demonstrated in this study, is expected to be valuable for ensuring the provision of high-quality data, particularly after the launch of new optical satellites in the future.