11:00 AM - 1:00 PM
[ACG38-P09] High resolution cloud motion vectors in GSMaP
Keywords:GSMaP, cloud motion vectors, Himawari-8
Global Satellite Mapping of Precipitation (GSMaP) is an hourly global precipitation map provided by JAXA, with a 0.1° grid between 60°S-60°N. Due to the effects of climate change, more and more typhoons and torrential rains, which cause localized meteorological changes in a short period, trigger many disasters worldwide. Therefore, there has been a growing need for GSMaP in the field of disaster prevention in regions where meteorological observation networks are not well developed. Since these meteorological phenomena vary locally in a short time, GSMaP is required to estimate them with higher temporal and spatial resolution.
In order to improve the spatial-temporal resolution of GSMaP, it is necessary to improve the resolution of cloud motion vectors which are used to estimate the movement of rainfall areas. This is because GSMaP employs precipitation maps obtained by microwave radiometers onboard low-earth-orbit satellites, which are not capable of global observations, and extrapolate rainfall over unobserved areas with the cloud motion vectors and the maps. The cloud motion vectors in the present GSMaP system are derived by pattern matching at one-hour intervals and 2.5° grid spacing, and then interpolated to 0.1° intervals. The higher spatial-temporal cloud motion vectors are required to estimate the movement of rainfall areas with higher accuracy.
Himawari-8, launched in 2014, provides higher-resolution infrared brightness temperature data every 10-minute. In this study, we use the high-resolution data of Himawari-8 to derive the high-resolution cloud motion vectors with a smaller grid and a shorter interval of 10 minutes. Then, we verify the accuracy of the cloud motion vectors by calculating the correlation coefficient and RMSE for the zonal and meridional components, respectively, by comparing with wind speed data in the Local Forecast Model GPV (LFM).
For the cloud motion vectors used in the current GSMaP, the correlation coefficient and RMSE are 0.48 and 10 m/s for the zonal component and 0.16 and 14 m/s for the meridional component, respectively. On the other hand, when employing an interval of 10 minutes and a grid interval of 0.6°-2.5° with the Himawari-8 data, the correlation coefficient and RMSE are 0.77-0.85 and 6.1-8.2 m/s for the zonal component and 0.64-0.70 and 10-12 m/s for the meridional component, respectively. For all the grid spacings, the results are more accurate than the current GSMaP. In particular, the maximum value of the correlation coefficient is 0.85 and the minimum value of the RMSE is 6.1 m/s when the grid spacing is 2.1°. In low accuracy areas, there are clouds extending at multiple altitudes on a single grid. In addition, we find that the cloud motion vectors derived with the grid intervals of 0.6° to 2.0° captured the movement of localized rain clouds better than those derived with the grid interval of 2.5°. In particular, 0.6-degree motion vector obtains detailed movement of clouds from occurring to development. In summary, the higher spatial-temporal cloud motion vector derived by the Himawari-8 data is more accurate than the current GSMaP.
In order to improve the spatial-temporal resolution of GSMaP, it is necessary to improve the resolution of cloud motion vectors which are used to estimate the movement of rainfall areas. This is because GSMaP employs precipitation maps obtained by microwave radiometers onboard low-earth-orbit satellites, which are not capable of global observations, and extrapolate rainfall over unobserved areas with the cloud motion vectors and the maps. The cloud motion vectors in the present GSMaP system are derived by pattern matching at one-hour intervals and 2.5° grid spacing, and then interpolated to 0.1° intervals. The higher spatial-temporal cloud motion vectors are required to estimate the movement of rainfall areas with higher accuracy.
Himawari-8, launched in 2014, provides higher-resolution infrared brightness temperature data every 10-minute. In this study, we use the high-resolution data of Himawari-8 to derive the high-resolution cloud motion vectors with a smaller grid and a shorter interval of 10 minutes. Then, we verify the accuracy of the cloud motion vectors by calculating the correlation coefficient and RMSE for the zonal and meridional components, respectively, by comparing with wind speed data in the Local Forecast Model GPV (LFM).
For the cloud motion vectors used in the current GSMaP, the correlation coefficient and RMSE are 0.48 and 10 m/s for the zonal component and 0.16 and 14 m/s for the meridional component, respectively. On the other hand, when employing an interval of 10 minutes and a grid interval of 0.6°-2.5° with the Himawari-8 data, the correlation coefficient and RMSE are 0.77-0.85 and 6.1-8.2 m/s for the zonal component and 0.64-0.70 and 10-12 m/s for the meridional component, respectively. For all the grid spacings, the results are more accurate than the current GSMaP. In particular, the maximum value of the correlation coefficient is 0.85 and the minimum value of the RMSE is 6.1 m/s when the grid spacing is 2.1°. In low accuracy areas, there are clouds extending at multiple altitudes on a single grid. In addition, we find that the cloud motion vectors derived with the grid intervals of 0.6° to 2.0° captured the movement of localized rain clouds better than those derived with the grid interval of 2.5°. In particular, 0.6-degree motion vector obtains detailed movement of clouds from occurring to development. In summary, the higher spatial-temporal cloud motion vector derived by the Himawari-8 data is more accurate than the current GSMaP.