Presentation Information
[AAS15-P08]Properties of a High-Resolution Global Atmosphere-Ocean Coupled Assimilation System as a Numerical Weather Prediction System
*Toshiyuki Ishibashi1, Yosuke Fujii1, Ichiro Ishikawa1 (1.Meteorological Research Institute, Japan Meteorological Agency)
Keywords:
data assimilation,weakly coupled data assimilaiton,Atmosphere-ocean coupled model,numerical weather prediction
Global atmospheric state estimation using advanced data assimilation systems, such as variational methods, can generate physically balanced, highly accurate atmospheric fields by integrating sophisticated global model predictions with vast amounts of observational information, such as satellite observations. However, the interaction between the atmosphere and ocean is only partially considered. Therefore, more precise handling of atmosphere-ocean interactions in data assimilation is expected to enable more accurate atmospheric and ocean state estimation and forecasting. Research on coupled atmosphere-ocean assimilation has progressed rapidly in recent years, but there are many challenges in estimating the state of the coupled atmosphere-ocean system, and thus far, improvements in atmospheric forecast accuracy through coupling have been limited (Browne et al. 2019). In this study, we develop a coupled atmosphere-ocean assimilation and forecasting system (MRI-CDA2) based on the Japan Meteorological Agency's operational atmospheric assimilation system, ocean assimilation system, and coupled atmosphere-ocean model, and evaluate its properties as a NWP system.
The coupled assimilation method of MRI-CDA2 is a weakly coupled assimilation method, similar to MRI-CDA1 (Fujii et al. 2021). The horizontal resolutions of the atmospheric and oceanic models are 20 km and 0.25 degrees, respectively, enabling direct performance comparison with operational global numerical weather prediction systems. The assimilation system uses 4D-Var for both the atmosphere and ocean. A one-month assimilation forecast cycle experiment using MRI-CDA2 was conducted, and the performance of the coupled assimilation system as an NWP system was evaluated by comparing it with an atmosphere-only assimilation forecast cycle. Improved forecast accuracy was observed for lower atmospheric temperatures, particularly in the tropics and Southern Hemisphere. This study did not apply the relaxation of the forecast field to the SST product, as was done in previous studies, and therefore can be said to have more directly incorporated the effects of atmosphere-ocean coupling. The main deterioration in forecast accuracy was observed in areas of reduced sea ice in the Arctic Ocean, indicating that assimilation, prediction, and handling of observation data for sea ice areas remain challenges for the future.
This research was partially supported by JSPS KAKENHI Grant Number JP24H02226.
The coupled assimilation method of MRI-CDA2 is a weakly coupled assimilation method, similar to MRI-CDA1 (Fujii et al. 2021). The horizontal resolutions of the atmospheric and oceanic models are 20 km and 0.25 degrees, respectively, enabling direct performance comparison with operational global numerical weather prediction systems. The assimilation system uses 4D-Var for both the atmosphere and ocean. A one-month assimilation forecast cycle experiment using MRI-CDA2 was conducted, and the performance of the coupled assimilation system as an NWP system was evaluated by comparing it with an atmosphere-only assimilation forecast cycle. Improved forecast accuracy was observed for lower atmospheric temperatures, particularly in the tropics and Southern Hemisphere. This study did not apply the relaxation of the forecast field to the SST product, as was done in previous studies, and therefore can be said to have more directly incorporated the effects of atmosphere-ocean coupling. The main deterioration in forecast accuracy was observed in areas of reduced sea ice in the Arctic Ocean, indicating that assimilation, prediction, and handling of observation data for sea ice areas remain challenges for the future.
This research was partially supported by JSPS KAKENHI Grant Number JP24H02226.