This study proposed a method to improve the geometric accuracy of the Advanced Himawari Imager (AHI), an instrument on board the recently launched Himawari-8 spacecraft, focusing on the data set over Japan acquired in the regional acquisition mode. The AHI scanning mechanism consists of four west-east horizontal sweeps, and derived four sub-images are combined into a single dataset. This scanning mechanism gives rise to variations in geometric accuracy among the four sub-images. These variations were quantified by analyzing the AHI band-3 images, which has the highest spatial resolution (0.5km) among the AHI bands. The geometric error, determined as the root mean square of average errors in each scan, was found to be approximately 0.3 pixels in the vertical (north-south) direction and 1 pixel in the horizontal direction during 8 hours of daytime. The geometric error showed two characteristics : temporal fluctuation, and variation by the scan. This study proposes a technique to improve these ‘scan-dependent' geometric errors by shifting the four sub-images independently based on the average error in each scan. The proposed method successfully reduced the geometric errors to 0.07 pixels in the vertical and 0.15 pixels in the horizontal direction, and hence improved the quality of time series composite images obtained using the AHI. Such geometrically-improved AHI image would enhance the accuracies of multi-temporal analyses used in the monitoring of vegetation, land use/land cover change, disasters and so on.
