2:30 PM - 2:45 PM
[AAS02-04] Typhoon seasonal forecasting by a high-resolution coupled GCM (NICOCO)
Keywords:typhoon, seasonal forecast, global model
To mitigate the impact of typhoons, it is needed to precisely predict typhoon activity before the beginning of the typhoon season (June). Some research institutes, operational centres, and insurance companies abroad issue seasonal forecasts of typhoons. In Japan, a part of private weather companies provides an outlook of typhoon activities, but no official seasonal typhoon forecast is issued by JMA.
Dynamical-based typhoon seasonal forecasts using conventional coupled GCMs has been intensively examined along with the progress of high-performance computers in the recent couple of decades. However, horizontal resolution is not high enough to represent observed typhoon intensity and some bias correction technique is needed to predict the intensity-related index (e.g., ACE) quantitatively.
Here, we used a 14-km-mesh global nonhydrostatic atmospheric model coupled with a 0.25-deg-mesh global ocean model (NICOCO; an AGCM NICAM coupled with an OGCM COCO) for 10-year (2010-2019) typhoon seasonal forecast experiments. The model is initialized on 20 May of each year and integrated to 1 November. The initial conditions for the atmosphere are made by interpolating the ensemble analysis data of ALERA (the first 16 members only are used because of computational resources). The initial conditions for the ocean are created by driving COCO using JRA-55do. Thus, the ensemble size of the experiments is five and the initial condition for the ocean is common among the ensemble members. We also performed NICAM (atmosphere-only) experiments. For NICAM experiments, the observational SST of OISSTv2.1 is given.
The results show that NICOCO performed better in predicting seasonal (June–October) ACE than NICAM. The correlation coefficients between simulated and observed seasonal ACE are higher in NICOCO experiments (0.79) than that in NICAM experiments (0.62). Both NICOCO and NICAM overestimate ACE but the absolute value of mean error is 61% lower in NICOCO than in NICAM. The model showed the best performance in predicting the seasonal number of typhoons in the eastern south part of the WNP (0°–18°N, 140°–180°E), where intense typhoons often form. The correlation coefficients between simulated and observed seasonal (June-October) numbers of typhoons are higher in NICOCO experiments (0.75) than that in NICAM experiments (0.71). For ACE in the eastern south part of the WNP, the correlation coefficient for NICOCO experiments is 0.86 and that for NICAM experiments is 0.71. These results demonstrate NICOCO’s good performance in typhoon seasonal forecasting.
Dynamical-based typhoon seasonal forecasts using conventional coupled GCMs has been intensively examined along with the progress of high-performance computers in the recent couple of decades. However, horizontal resolution is not high enough to represent observed typhoon intensity and some bias correction technique is needed to predict the intensity-related index (e.g., ACE) quantitatively.
Here, we used a 14-km-mesh global nonhydrostatic atmospheric model coupled with a 0.25-deg-mesh global ocean model (NICOCO; an AGCM NICAM coupled with an OGCM COCO) for 10-year (2010-2019) typhoon seasonal forecast experiments. The model is initialized on 20 May of each year and integrated to 1 November. The initial conditions for the atmosphere are made by interpolating the ensemble analysis data of ALERA (the first 16 members only are used because of computational resources). The initial conditions for the ocean are created by driving COCO using JRA-55do. Thus, the ensemble size of the experiments is five and the initial condition for the ocean is common among the ensemble members. We also performed NICAM (atmosphere-only) experiments. For NICAM experiments, the observational SST of OISSTv2.1 is given.
The results show that NICOCO performed better in predicting seasonal (June–October) ACE than NICAM. The correlation coefficients between simulated and observed seasonal ACE are higher in NICOCO experiments (0.79) than that in NICAM experiments (0.62). Both NICOCO and NICAM overestimate ACE but the absolute value of mean error is 61% lower in NICOCO than in NICAM. The model showed the best performance in predicting the seasonal number of typhoons in the eastern south part of the WNP (0°–18°N, 140°–180°E), where intense typhoons often form. The correlation coefficients between simulated and observed seasonal (June-October) numbers of typhoons are higher in NICOCO experiments (0.75) than that in NICAM experiments (0.71). For ACE in the eastern south part of the WNP, the correlation coefficient for NICOCO experiments is 0.86 and that for NICAM experiments is 0.71. These results demonstrate NICOCO’s good performance in typhoon seasonal forecasting.