Abstract
Comparative observations between two or more sensors viewing the same target(s) of known or stable radiance close in time have been successfully used for decades to calibrate and cross-compare calibration of various sensors. While the common target approach has proven effective in ensuring data consistency, they typically involve coordinating near-simultaneous data acquisition that are resource-intensive. The need to address geometric, spectral, atmospheric, and angular differences between measurements adds further complexity, making these methods challenging to implement on a routine basis. This study uses an approach to evaluate radiometric scales of two sensors using statistics of their global coverage of the Earth surface as the common target. Hence, it is agnostic to time, site, atmospheric condition, or angle considerations. We demonstrate feasibility of that concept by assessing the level of agreement between Landsat 8 and Landsat 9 reflectance data using two variations of the approach, namely the Data Average and the World Average methods. The primary goal is to evaluate the consistency of the reflectance measurements from the two satellite datasets across 16-day periods. We use metrics such as the ratio of as-observed reflectance values and Earth Mover’s Distance (EMD) to evaluate differences between the two datasets and describe how to mitigate the risk of potential oversampling when comparing global data. Results of the World Average and Data Average methods are consistent with the previous Landsat 8 and Landsat 9 underfly studies, underscoring excellent agreement between the radiometric scale of the sensors onboard the two platforms.
World Average Intercomparison Method: An Approach to Compare Landsat 8 OLI to Landsat 9 OLI
Comparative observations between two or more sensors viewing the same target(s) of known or stable radiance close in time have been successfully used for decades to calibrate and cross-compare calibration of various sensors. While the common target approach has proven effective in ensuring data consistency, they typically involve coordinating near-simultaneous data acquisition that are resource-intensive. The need to address geometric, spectral, atmospheric, and angular differences between measurements adds further complexity, making these methods challenging to implement on a routine basis. This study uses an approach to evaluate radiometric scales of two sensors using statistics of their global coverage of the Earth surface as the common target. Hence, it is agnostic to time, site, atmospheric condition, or angle considerations. We demonstrate feasibility of that concept by assessing the level of agreement between Landsat 8 and Landsat 9 reflectance data using two variations of the approach, namely the Data Average and the World Average methods. The primary goal is to evaluate the consistency of the reflectance measurements from the two satellite datasets across 16-day periods. We use metrics such as the ratio of as-observed reflectance values and Earth Mover’s Distance (EMD) to evaluate differences between the two datasets and describe how to mitigate the risk of potential oversampling when comparing global data. Results of the World Average and Data Average methods are consistent with the previous Landsat 8 and Landsat 9 underfly studies, underscoring excellent agreement between the radiometric scale of the sensors onboard the two platforms.