5:15 PM - 7:15 PM
[ACG41-P08] A Methodology for AMSR2 Multi-Pass Resolution Enhancement Over Sea-Ice Covered Oceans
Keywords:AMSR2, Resolution enhancement, Denoising, Sea Ice
Satellite passive microwave radiometers can observe the Earth's surface globally every day, regardless of nighttime or cloud cover, and have been widely used to estimate sea ice physical parameters, such as sea ice concentration. However, their spatial resolution (ranging from a few to several tens of kilometers) is often insufficient for capturing surface complexities in sea ice regions. Several studies have attempted to synthesize higher-resolution brightness temperatures (TBs) by utilizing the sensor antenna pattern. Among these approaches, the microwave version of the Scatterometer Image Reconstruction (rSIR) method, developed by Early and Long (2001), effectively enhances spatial resolution at a low computational cost by integrating multiple passes per day in polar regions. However, this method assumes that the TB field remains unchanged between satellite passes.
For the Global Change Observation Mission-Water/Advanced Microwave Scanning Radiometer 2 (GCOM-W/AMSR2), the minimum time interval between two passes over a given location is approximately 100 minutes. During this period, sea ice motion and atmospheric variability, such as changes in water vapor and cloud cover, cause variations in TB at each grid cell, which degrades the performance of rSIR. To address this issue, we have developed a methodology for multi-pass resolution enhancement to mitigate such performance degradation.
In this study, we focused on enhancing the resolution of TBs at 19 GHz, 36 GHz, and 89 GHz from the GCOM-W/AMSR2 Level-1B product. First, we applied the unidirectional total variation minimization method proposed by Bouali and Ladjal (2011) to remove stripe noise from the 89 GHz TBs. Then, for the variability in atmospheric influence between passes, we introduced the following adaptive approach: in areas where TB differences between two passes are large and spatially extensive, single pass rSIR is applied to minimize the influence of atmospheric variability. Conversely, in regions with minimal or small-scale TB differences, multi-pass rSIR is employed to enhance spatial resolution. Specifically, we applied total variation L1 regularization and a median filter to identify regions where multi-pass rSIR should be applied. Subsequently, considering the impact of sea ice motion on TB variation, the Lucas-Kanade method was used to estimate subgrid-scale sea ice displacement between passes, thereby ensuring precise coregistration of the two passes. These preprocessing steps effectively reduced inconsistencies in TB between passes and enabled the reconstruction of more accurate high-resolution TBs. In this presentation, we demonstrate examples of our enhanced-resolution TBs and compare them with MODIS data with a higher spatial resolution.
For the Global Change Observation Mission-Water/Advanced Microwave Scanning Radiometer 2 (GCOM-W/AMSR2), the minimum time interval between two passes over a given location is approximately 100 minutes. During this period, sea ice motion and atmospheric variability, such as changes in water vapor and cloud cover, cause variations in TB at each grid cell, which degrades the performance of rSIR. To address this issue, we have developed a methodology for multi-pass resolution enhancement to mitigate such performance degradation.
In this study, we focused on enhancing the resolution of TBs at 19 GHz, 36 GHz, and 89 GHz from the GCOM-W/AMSR2 Level-1B product. First, we applied the unidirectional total variation minimization method proposed by Bouali and Ladjal (2011) to remove stripe noise from the 89 GHz TBs. Then, for the variability in atmospheric influence between passes, we introduced the following adaptive approach: in areas where TB differences between two passes are large and spatially extensive, single pass rSIR is applied to minimize the influence of atmospheric variability. Conversely, in regions with minimal or small-scale TB differences, multi-pass rSIR is employed to enhance spatial resolution. Specifically, we applied total variation L1 regularization and a median filter to identify regions where multi-pass rSIR should be applied. Subsequently, considering the impact of sea ice motion on TB variation, the Lucas-Kanade method was used to estimate subgrid-scale sea ice displacement between passes, thereby ensuring precise coregistration of the two passes. These preprocessing steps effectively reduced inconsistencies in TB between passes and enabled the reconstruction of more accurate high-resolution TBs. In this presentation, we demonstrate examples of our enhanced-resolution TBs and compare them with MODIS data with a higher spatial resolution.